What AI search can tell you about your content strategy
See which client questions your content isn’t answering, then decide what to update, combine or create.
Most firms now have some evidence about how their content performs – but turning it into clearer priorities is harder. A content audit or AI visibility score on its own does not tell you which gaps matter to clients, or where the next piece of thought leadership should go.
The most useful evidence shows where people cannot find a clear, complete or credible answer. Combined with client insight, market evidence and expert judgement, it becomes the basis for a stronger content and thought leadership strategy.
In this session, Meridian West and Squiz discuss what discovery evidence can reveal, which gaps are worth acting on and how to build a continuous planning loop rather than a one-off report.
Watch the Webinar
Transcript: Watch the Webinar
Alastair: Welcome, everyone, to the second of three conversations with the Meridian West and Squiz teams, exploring AI search, thought leadership and content, and how things are changing.
In our second conversation today, I’m delighted to be joined yet again by Ethan and Will. How are you guys doing today?
Speaker: Yeah, good, Alastair.
Alastair: Great. And a new voice in today’s conversation – Jamie. How are you doing, Jamie? Coming to us from Madison, Wisconsin.
Jamie: Yes. Hi. Thanks for having me, Alastair. Great to be here and really looking forward to the conversation. I’m the vice president of partnerships here at Squiz and work with partners like Meridian West and others. I’m really just here to help solve problems for our customers.
Alastair: Great. Great to have you on today’s conversation.
So what are we looking at today? As I said, this is the second of three conversations, where we’ll be looking at what AI search can tell you about your content strategy. In particular, we’re going to be looking at some practical ideas around spotting gaps in content, what AI search can tell you about your content, and how you can use that insight to create a virtuous loop to decide what you improve, consolidate or create in terms of new content.
So, excited to dive into all of this. Before we do, Will, do you want to give the audience a quick reminder on Squiz?
Will: Absolutely. “Virtuous” is a great word – I’ll have to take that one.
Sure. So firms like Gadens, Thomsons and many others use the Squiz platform to build websites that are more than a glorified brochure – so personalisation, multivariate testing and AI search experiences, all included.
But also, increasingly, we monitor our clients’ content for content health, identifying contradictions, tone of voice and policy adherence.
Alastair: Fantastic. And for those of you who aren’t familiar with Meridian West, we work across professional services and B2B organisations. One of the key areas that we support is content and thought leadership, helping firms with their content strategy and development as well. So, great to be partnering with Squiz.
As I said, this is the second of our webinars. Those of you who joined us a couple of weeks ago will remember that we were talking about thought leadership, why it matters and how it’s important to the buying decision.
We were also talking, Will, a little bit about the changing nature of AI search as well. Do you want to remind those who perhaps missed the webinar what they missed last time?
Will: Yes. So, unless you’ve been under a rock, obviously AI search has changed discovery and how owned content can be missed. We talked about what can happen when your content isn’t cited. If you’re burying thought leadership in PDFs, you’ve got very vague titles or you don’t have the right technical standards, that can certainly hurt you.
I think we also ended by looking at the four signals that decide whether you’re mentioned or cited in AI search: structure, consistency, authority and freshness.
So definitely worth a revisit – but obviously after this webinar. Don’t leave now.
Alastair: Absolutely. The link is in the chat, so don’t click on it yet, but you can click right at the end of this conversation and go back and see what you missed if you didn’t attend.
We’ve got a third conversation coming up and we’ll preview that right at the end today. We do want to make this interactive, so if people have got questions or ideas, put them in the Q&A or the chat. We’ll take a look at those as we go through.
But we want to start, as we did in the previous conversation, with a bit of an audience poll. Hopefully you can see that on the screen – it’ll pop up for you in a minute.
We’re asking you how you currently assess the impact of your thought leadership content. It’s a big topic that a lot of firms we’re talking to at the moment are really trying to understand: impact and return on investment. So we’ll give people a moment to fill in some thoughts there.
Will: And of course it’s all a little bit up in the air, right? The forms and units of measurement that we’ve relied on over the last 15 years aren’t so salient anymore, which is fun and interesting, but also terrifying at the same time.
Alastair: It is, and I think we’re all learning this together, right? The world is evolving and we’ve got to adapt and evolve. I think content is a big part of that evolution.
We’re seeing the numbers bounce around a little bit. We might just leave that on screen for another few seconds.
Jamie, Ethan, anything jump out at you from the results we’re seeing so far?
Jamie: Yeah. To me, I’m actually surprised that there’s not any on the second point: gathering feedback on whether clients find it useful.
I think that’s a really important area to focus on – asking clients, “What do you want to learn? What do you want to understand about our particular firm?” or whatever that may be. So honestly, I find it surprising that that’s not bouncing around a little bit.
Alastair: Yeah. Ethan, any thoughts?
Ethan: Yeah. I guess just under 20% at the moment, assessing the impact of it. And I think that’s something we’ll talk about today: how do you start to use AI search, and other tools as well, to take the first step, which is to evaluate how a piece of thought leadership or content lands?
It seems like a lot of them are doing that, but I still think there is a bit of a demographic who are struggling to get off the ground.
And interesting that a third are assessing it by looking at new business. When we speak to our clients, the goal is often how they can link content back to new business and sales. But that’s a tricky one, right? Tricky to measure and tricky to understand.
Alastair: Absolutely.
So let’s frame the conversation a little bit. We’ve been talking about AI search – a big topic for our last conversation. I guess a question to you, Will, and Jamie, your thoughts as well: what can this tell us about firms’ existing thought leadership content?
Will: Yeah, sure. So I think I alluded to it at the beginning: AI search has obviously changed the discovery-to-decision process.
Historically, what thought leadership or insights content was there to do was to educate, inform and give people an understanding of a salient topic. Through that process, through osmosis, you then learn more about the brand and the services that that organisation can provide. And of course now all of that discovery process is happening outside of your own domain.
But thought leadership content does still influence recommendations to your firm because it’s the foundational information which the models give to end users. A mention in an answer, which is actually new for post-SEO, can be positive or negative – like, “Don’t go speak to AdultShore because they’re not great at litigation”, right?
And a mention doesn’t always guarantee a citation, so it might not even have a link to your firm, and they’ve used your IP anyway. You can also be cited in an answer, but you’re not actually mentioned by name in the body of the response at all.
So there are all these interesting nuances that we’re looking at. Here at Squiz, what I think we’ve found is that AI search does expose a lot of bad habits and bad content habits that we’ve been able to brush over, which have now been thrown up to the surface.
And so, to answer this, we developed some tooling that can identify some of these bad habits.
If we jump to the next slide.
Speaker: Yeah, here we go.
Jamie: Yeah, and to your point, Will, I think the bad habits – it’s scary, right? Because it’s like, “Oh gosh, all this is coming out now and we have to maybe revamp and revisit all of our existing content and create new content.”
But I think that comes down to what you can actually control on your site and what you can’t control. You can’t control what someone’s going to search for on AI. You can’t control exactly what that’s going to serve up every single time because it’s so nuanced to that specific user and that specific search pattern.
But what you can control is what’s actually on your site and the content – making sure it’s credible, it’s reliable, it’s not fluff, it’s niche-focused where it needs to be, it’s specific where it needs to be. All of those things really matter.
Going back to the poll, I find it interesting that the highest, at 50%, was reviewing content performance to decide where to improve, and I think that’s a really good area to start.
I think these are maybe some ideas of where to look and what AI has uncovered, so what you can improve. Would you agree with that, Will, on these five points?
Will: Yeah, absolutely.
That ties into how we monitor performance and those bottom ones about: does our content clearly answer these questions that we want to be cited in? I think we’ll talk about that later on in this talk. And does it answer the questions that we know our audience are getting at?
Some of that top-level row there is, like you said, the foundational stuff of bad habits. If we can fix those, that improves the likelihood of our thought leadership reaching people.
If you have outdated information on your site or you contradict yourself, that’s what the models get tripped up on and that’s when they start hallucinating, or they most likely just use another source. If you contradict yourself about an hourly rate, for example, in one part of the site compared to another.
And when you’re working with experts in a particular field, they’re predisposed to use lots of technical jargon and they want to write to show off their high-level intellect, which isn’t necessarily as accessible for human readers. But the models are in the same boat. If you can clearly spell out your point of view in clearer language, that increases your mention as well.
Alastair: And I think it’s interesting, isn’t it? We’re in a position now where firms have got more data than ever about what’s visible, what’s not visible – the things that both of you have been talking about.
Yet, Ethan, I think we still find in a lot of the conversations we have, it’s really about firms knowing where to start and where to put that investment into content development in the first place.
I’d be interested in your thoughts and experiences of how firms are spotting those gaps and creating new ideas, because the world is changing all the time, right? Not just the technology, but the reach, the scope, the ideas and the themes that firms seem to be talking about.
Ethan: Yeah, 100%. I think we can probably talk about this first-hand. We work with a lot of firms to help them create thought leadership and content, from the initial design to actually creating the report and the outputs.
A lot of time is spent upfront because it is really hard to find a distinctive topic or an area that you want to cover that you feel is not just reaching your intended target audience but also – and you can see by the chart there – finding this sweet spot between the audience and somewhere where you can have a point of view that’s going to differentiate your content from other content out there.
But it’s also going to link back to what you’re trying to do, what dial you’re trying to move from a commercial sense: are you trying to drive new sales in a new market or move some brand needles?
Some of our ATLAS research – for those who are new to this session, it’s research that we’ve done on approaches to thought leadership – shows that respondents find it hard to find distinctive topics and areas to cover. I think there’s just a lot of content out there and everyone’s getting quicker to get into market and talk about stuff.
As you mentioned, when we do a lot of programmes, we do ideas labs and brainstorming sessions, considering these different factors to try to get to both your internal view – what you want to say – alongside the client view.
Then also aligning that with a strategy to get, maybe not necessarily one area for one campaign, but a year’s worth of strategy. What are the different things that you want to say over that spell?
If we go on to the next slide, we pull out some ATLAS insights where we asked respondents where their ideas for topics came from over the last year. As we can see, there’s a mixture of different internal, market and client sources.
One of the things that stands out is that ideas still predominantly come from internal partners and directors, probably with an element that they want to say something about what they’re doing or promote their expertise, which I think is really valuable. You want that level of engagement internally.
But often that process can be quite ad hoc, quite raw. Partners come with different ideas, they can be quite contradictory, and it just means that some of your content on a wider scale, stepping back, can be quite messy.
One voice that I think this chart shows is slightly underrepresented is the client. Coming back to what you mentioned, Jamie, getting that anecdotal feedback from clients on not only what content was impactful and worked for them, but also what challenges they’re having. What things are on the horizon for them?
Not many firms do it, but some firms are starting to integrate their client listening programmes as a source of insight.
I think all the firms that we work with – large professional services firms – will have some sort of client listening campaign, but not many integrate detail from that into, “OK, what content are we going to do? What are clients telling us that they’re looking to do over the next 12 months?”
Some firms ask that quite directly. They would simply ask, “What content would you like to see from us?” or “What would you like to read or view over the next year?”
Some ask it a little bit more softly: “What are your challenges? What are you trying to do over the next 12 months?” But either way, I think there’s quite a big source of information that you’re missing there.
It makes up our sweet spot diagram: integrating what you want to say, what your clients are looking to hear, but also what’s going to help you move the dial on a commercial goal.
Will: Ethan, just a question on that point. Do you work with organisations that do any kind of social listening, or is that primarily a B2C initiative?
Ethan: I think it’s generally a bit more B2C. We do work with a few firms who do wider – I guess it’s more B2B – brand work, where they’ll look at the performance of content alongside some wider global brand activity and see how that’s going to change in a tracker. But I think generally less so in the professional services realm.
Alastair: Yeah. I think there’s a great opportunity there, right? I guess that’s why you’re asking the question?
Will: Yes, because it basically takes an aggregate view of those client sources and uses the latest in technology to try to understand what it is that people in the market are talking about that could influence the content you create.
Alastair: Yeah. And I think that’s where, certainly, those larger firms – if you’re a Big Four firm, if you’re McKinsey – they are doing a lot of that already. But I think it becomes harder for smaller, mid-market firms.
It goes back to everything that Ethan’s just been talking about, right? There are lots of competing interests internally. You need to overlay that with the firm’s priorities and where you can afford to make these investments, but also what data and intelligence you can get from the work Squiz is providing on that.
So, tying some of these threads together, what we want to think about for our audience today is: how do you tie this together to make some of those decisions?
Because you’re probably facing, looking ahead to 2027, thinking about what do we need to do more of, what do we need to do less of, where do we need to improve? What’s the way through this quagmire, Ethan? What do you think?
Ethan: Yeah. Well, again, I think it’s tricky.
From our experience, whether at the start of a campaign or whether it becomes more of an ongoing thing, the best thing to do is to try and get all the people involved in that process into a room, into a workshop – especially if the ideas are going to be stemming from different partners and fee earners.
People doing stuff in silos and pockets is probably the most unhelpful process, just because you create different content and maybe cover some things too much or don’t cover others enough.
When we do this well with our clients, whether it’s for an individual campaign or looking at the year ahead, we try to get all those decision-makers in a room for a few hours and bring in their ideas. As we see, people generally have ideas for content they want to create.
But we also bring in those external voices. We’ll use client listening or more anecdotal feedback. We’ll do some desk research, maybe using a bit of AI to help us see what’s in the market.
And then also looking back and saying, “OK, looking over the last 12 to 18 months, how impactful was your content?” In terms of downloads, engagement with posts, but also maybe AI performance as well.
Then we’re filtering all these different ideas and channels to come up with, whether it’s for one individual campaign or plotting over the next year, the content that we’re looking to cover.
It is a challenge because there are a lot of different sources and everyone’s got ideas. But I think the best way is to get them in the room together and then come up with something which has variety but is also strategic.
Without going back to the chart before, I think one of the big things was that less than half actually considered the growth strategy when coming up with content. I felt that should be 100%, really, right?
Depending on what you’re trying to do over the next year, that should be a massive factor in the thought leadership you’re producing. It’s a really powerful tool for changing perceptions of the firm, or driving growth or new sales. So it should be really aligned: “OK, let’s get it all under a strategy and make this a strategic pull from everyone.”
What’s your take, Will?
Will: Yeah, well, my cogs are whirring, because you mentioned earlier, Alastair, about if you’re not the Big Four and you’ve got a limited budget, what are your options?
One of the most amazing things about AI is the persona work that you can create, right? You can create your gems or your plugins now that work as a particular persona. If you’re in BD and marketing, you can then get it to devise hypothetical questions about what they are asking, rather than always having to go internally and speak to these experts.
It’s not perfect. But what it does allow you to do is generate a bunch of questions that a potential client could be asking their models, and then you can go and run a test over your own content.
If I was going to ask a question of that calibre against our content, does my website answer it? Can it accurately answer it? Can it do it clearly? And if not, then there’s your gap, right? There you can go and write content or put that into your content campaigns and schedule to look at it.
We also work with lots of clients on internal search – that search bar on your website. We do a lot of work building generative experiences for clients there. When you allow visitors to ask long-form questions against your own content, that is first-party data that’s invaluable, because people are asking questions of you.
They come to your site, so that’s a smaller cohort, obviously, than in the big LLMs. And if they’re asking questions of your content and it doesn’t give a direct response, that can inform where you need to write content in the future.
From that, we’re actually finding a lot of process questions, like, “What’s it like to engage with you? Can you walk me through a current engagement or a matter?”
That’s what we want to know just before we click and hit “Contact us”. We haven’t historically had to write content for that, so a lot of our firms have uncovered that.
Alastair: Yeah, that’s fascinating to me. The idea that people are asking these experiential questions, right? It’s not just the factual capability questions, it’s actually, “Tell me more about the experience, what I can expect.”
And as you say, if that content isn’t there in the first place, there’s nothing that’s going to come back when you’re asking those questions. So it’s about being really thoughtful about that.
I think that ties into what you were saying earlier, Ethan, about increasingly overlaying insights from clients directly onto this content strategy as well.
I know we had a second poll question that we wanted to ask people, so why don’t we move on to that one?
We’ve got one which should be coming up for you to answer now: what’s your biggest barrier to ongoing content improvement? I know there are probably quite a few that people may be facing. You might have to think hard between the choices, but what’s the biggest one that you’re facing?
We’ll leave that up for a few seconds as we’re getting some answers in.
Jamie: While that’s up, Alastair, I was going to say something back to Ethan’s point, and this goes with clarifying who is responsible.
Ethan, you mentioned having all the decision-makers in a room and then bringing the results from all of the assessments that you’ve done and showing that.
I’m curious: do you find that the decision-makers in those rooms are saying, “Oh, this is exactly what we expected to see”? Or do you find that it’s, “Oh, we didn’t even know this”, or “We didn’t even think about that”? Or is it maybe a mix?
Ethan: Maybe a bit of a mix. Maybe to Will’s point, often it’s like, “Oh, we know people are interested to hear about that”, but maybe, “Oh, that’s actually really specific”, or “That’s more about process rather than big legal trends.”
It’s more, “How is this specific legislation change going to impact my business here?” So, “Oh, that’s a very niche bit of thought leadership to do.”
But generally it’s more about consolidating their ideas. Often the ideas they bring in a very raw format are very close to what clients are asking. It’s just a sense of how do we reframe that so what you’re doing is answering the question that they’re trying to ask and find the answer to?
So I’d say often there is a lot of overlap there. It’s not, “No, we have no idea what this is or why we’d feel comfortable producing content around this.” It’s more the phrasing of clients and prospects – it’s quite interesting to see what they’re asking to actually read.
Alastair: Yeah. It’s interesting getting results in, and there seems to be quite a strong consensus here that the biggest barrier is about time.
I think that’s something that I see a lot: firms are in execution mode all the time. We need to deliver, we’ve got a brief, we need to deliver particular content, particular campaigns. There’s very little time for reflection – it’s on to the next thing.
I don’t know, Will, does that surprise you, what we’re seeing here, or does this resonate?
Will: Not one bit.
We talk about what AI has given marketing teams. It’s not less work, it’s more work, isn’t it? More demands for more content, more demands for more in-person events. Giving yourself space to be retrospective about, “Are we on the right course?”
I mean, Jamie and I can say that here at Squiz we’re probably in exactly the same boat.
Alastair: Jamie, what are your thoughts?
Jamie: Yeah, I think – and I’ll talk about this in a little bit – even finding the right time and the right people to do that with, and then what to improve next.
The top one, no one commented on, but I think they probably didn’t comment on it because sometimes you don’t know what to improve next. So it’s, “How do we move forward with that?”
I think having the right tools and the right assessments that Meridian West provides to help your next improvements is an easier way to go ahead versus just trying to guess.
Yeah, time, to your point, Will and Alastair, is always the hardest thing, isn’t it, for all of us?
Alastair: Absolutely. And time is marching on, and I want to make sure we leave a little bit of time to talk about continuous improvement, because that’s been a big theme of what we’ve been talking about today: building that into your everyday workflows.
So, a quick reflection from you, Ethan: what does ATLAS research tell us about this and the time horizons people are thinking about when it comes to content strategy?
Ethan: Yeah, well, I think it’s what you were just saying. It feels like most firms take quite a mixed approach, but the majority are focusing on shorter-term campaigns or content strategies that are typically wrapping up within a year or maybe within three months.
And I think that makes sense.
Are kind of hot off the press. Firms want to keep content up to date. They want to talk about the newest thing.
Not in all instances, but sometimes there’s a level of campaign fatigue and they do all this work to release the report and move on to the next thing.
But I think the most sustainable and most successful approach is creating this continuous loop of content, where you don’t have to reinvent the wheel all the time. You can take a lot of learnings from previous content and build upon pieces rather than having to create something from scratch all the time.
And I think that is a challenge, but that’s what the firms we’re seeing having most success in this instance are doing.
Alastair: Yeah. And Jamie, in terms of bringing this to life with that idea of the continuous improvement loop, tell us about how you see that working in practice.
Jamie: Yeah, as we all know in marketing, it’s not “set it and forget it”. You have to continuously improve, continuously repeat and find what’s working.
To Ethan’s point, start with: what are those short-term campaigns and those long-term campaigns, and what does that look like?
This is our framework. We actually used this on the Squiz website when we revamped it. We utilised our tool to help us with that.
First is listen, of course. Listen to what’s happening on the digital experience side, but also listen to your internal folks. Do some customer reviews. Ask customers, gain feedback. Maybe have a select group of great customers that you can leverage and lean on to help you with that and really find what that feedback is.
Assess, of course: what’s working, what’s not working? I think also with assess, it’s who’s doing that for you. To Ethan’s point earlier about the decision-makers, have clear ownership of who’s helping in that area.
Then, once you understand what you need to improve next, actually do it and execute on that, on your experience.
Review. Ask those folks again – maybe that select group of customers, your internal folks. Look at the data that the tooling provides, and then repeat what is working.
Obviously, repeat what’s not working. Go back to the first part of the framework: let’s listen and assess again.
We did it ourselves on the Squiz website and we found some gaps in our content that we could fix immediately, definitely for the short term. Then there are some that are going to take a little bit more time to go through later.
But I think it’s really important to do this. It is a continuous loop. It never ends. So finding the time is probably the hardest piece, but making sure you’re actually doing it is the biggest area.
Alastair: Yeah, absolutely. We wanted to leave you all, just before we wrap up, with a few practical takeaways from our conversation today.
One that really struck me was, as Ethan was saying, just under half of firms are really linking that content strategy to the broader firm growth strategy. Clearly a missed opportunity there for many. So make sure you’re really focusing on what’s going to deliver commercial value to the firm, ultimately what’s supporting that growth journey.
We’ve talked today about lots of the insights, data and intelligence that you can use to help inform that continuous improvement, making sure you’re overlaying those together.
And then lean into that continuous improvement loop. Jamie was just talking through how that works in practice at a high level, but it’s really important that that becomes part of your everyday thinking and you’ve built a strong workflow around that.
That leads us into the topic for our third conversation, coming up on 9 December, where we’ll be diving into that in a little bit more detail, and we’d love to have you join us for that.
But all that leaves me is to say thanks, Jamie, thanks, Ethan, thanks, Will. Great to be chatting to you today.
The link for the third webinar is in the chat as well, so do register for that. If you didn’t see part one, you can go and click away and watch back part one as well. We’d love to see you again on 9 December. So thanks, everyone.
Speaker: Thanks, Alastair.
Speaker: Thanks, guys. Bye.
Speaker: Bye.
Video: Watch the Webinar. Captions and transcript available on playback.
Poll Results

- Reviewing content performance to decide what to improve - 50%
- Tracking enquiries or new business linked to it - 33%
- We do not currently assess its impact - 16%
- Tracking whether it reaches the right people - 0%
- Gathering feedback on whether clients find it useful - 0%
- Finding time for regular reviews - 71%
- Measuring whether changes are working - 29%
- Knowing what to improve next - 0%
- Clarifying who is responsible - 0%
Watch the Europe webinar
Transcript: Watch the webinar (UK)
1
00:00:13.970 --> 00:00:22.069
Toby Margetts: Yeah, yeah, hi for those of you who've joined, just give it another, kind of, 30 seconds or so before, before we jump in.
2
00:00:39.320 --> 00:00:47.570
Toby Margetts: Alrighty, hello, and welcome, everybody. Thank you very much for, for joining us this morning. Really great to have you, have you on board.
3
00:00:47.760 --> 00:01:05.749
Toby Margetts: Today, we're going to be talking about how we can keep our content healthy in the world of AI. So, most content teams, I think, I'm pretty sure, I know, know what good content looks like and what it should do. It should be useful, it should be accurate, easy to find, easy to maintain.
4
00:01:05.770 --> 00:01:16.100
Toby Margetts: The hard part is keeping it that way once things like requests and updates and ownership changes and business priorities start moving faster than the team can realistically track.
5
00:01:16.650 --> 00:01:32.030
Toby Margetts: My name is Toby. I'm one of the Digital Strategy Directors here at Squiz. I'm joined by the wonderful Jamie Sharp, who is our Head of Customer. And just before we kind of jump in, I'm going to hand over to Jamie, who's going to say just a quick word on Squiz. So, Jamie, over to you.
6
00:01:32.030 --> 00:01:42.480
Jamie Sharp: Thank you, Toby, and morning, everyone. It's really nice to see so many new and familiar faces. So, just a little bit of background about Squiz for those of you who don't know us yet.
7
00:01:42.510 --> 00:01:58.999
Jamie Sharp: Squiz is a digital experience platform, so we've got a really powerful website platform that you can use to build websites, intranets, multiple sites, and get really great data out of that platform to help you prioritize how you drive your commercial marketing strategies. So, our platform is made up of three tools.
8
00:01:59.000 --> 00:02:14.070
Jamie Sharp: It's the digital experience platform, it's funnel-back conversational search, which brings that experience of AI search onto your website today, and then also content intelligence, and that will scan your website for AI search visibility, and it'll tell you exactly how to improve
9
00:02:14.070 --> 00:02:28.240
Jamie Sharp: to be found and to avoid that scenario of your competitors being found before you. And it's that AI visibility and the learnings and insights we've made about content from doing this ourselves and from working with hundreds of customers globally that's going to be informing what we're talking about today.
10
00:02:29.170 --> 00:02:31.469
Toby Margetts: Lovely stuff. Thank you very much, Jamie.
11
00:02:31.840 --> 00:02:49.169
Toby Margetts: So in terms of what we will be covering today, there are really four key things, I would say. Number one is why good content practice quietly breaks down. So I think with the best will in the world, it can be really tough to stay on top of your content's health for a number of reasons, and we'll get into that very shortly.
12
00:02:49.170 --> 00:03:01.640
Toby Margetts: Number two is the four AI discovery signals. We have actually talked about them in a previous webinar. Just as a quick reminder, we're talking about structure, metadata, authority, and freshness.
13
00:03:01.640 --> 00:03:15.380
Toby Margetts: keeping these healthy is absolutely essential. Number three is how to use those signals in a content discovery loop. What practical steps can you actually take to stay on top of ensuring your content adheres to those signals?
14
00:03:15.460 --> 00:03:26.730
Toby Margetts: For how Squiz's content intelligence tool actually helps you do this, even when you're dealing with thousands of pages of content, which I think it's fair to say most of our customers are.
15
00:03:28.080 --> 00:03:43.459
Toby Margetts: So, first up, why does good content practice frequently break down? So, none of this is new to a good content team. You already know content needs to be clear, accurate, useful, current, findable, and, of course, owned.
16
00:03:43.500 --> 00:03:56.450
Toby Margetts: The problem is, I think, what happens after publication, so new requests keep arriving, old pages keep on aging, similar answers appear in different places, ownership changes, audits go stale.
17
00:03:56.450 --> 00:04:04.460
Toby Margetts: over time, good practice breaks down, because no one actually has a current picture of the whole estate. That's a really difficult thing to do.
18
00:04:04.750 --> 00:04:12.900
Toby Margetts: A couple of examples, so, for example, there might be a policy update made on one part of the site, but it's not updated somewhere else.
19
00:04:12.960 --> 00:04:25.670
Toby Margetts: Maybe there's a PDF that's remained live after the web page it's been generated from has changed. Lots of ways that this can happen, and we don't often realize that we have these content issues until a user runs into them.
20
00:04:25.670 --> 00:04:34.499
Toby Margetts: And what I would say is that's kind of always been a problem, but AI is really putting a kind of enormous magnifying glass over these issues.
21
00:04:35.830 --> 00:04:53.980
Toby Margetts: I think a really good way to think about it, is that AI search is basically raising the cost of unhealthy content to organizations. So, I often find there's a bit of a misconception when it comes to content and AI. It's the idea that content is somehow sort of less important because of AI.
22
00:04:54.010 --> 00:05:02.499
Toby Margetts: But I think it's, it's actually the complete opposite, right? So, if the content estate contains duplicates, stale, or conflicting answers.
23
00:05:02.630 --> 00:05:18.490
Toby Margetts: AI systems have a much weaker foundation to work from. The really controllable part, I think, is not guaranteeing a citation, for example. The controllable part is actually making your content clearer, more current, and ultimately much more answerable.
24
00:05:19.590 --> 00:05:39.130
Toby Margetts: We're actually just going to jump into, a quick pile, bit of, early audience engagement, and I'm really genuinely interested to see what comes up here. It's gonna be, it's going to be interesting. But the question is, and we'd love everybody to answer if they can, which of these do you see most often in your content? So…
25
00:05:39.130 --> 00:05:50.660
Toby Margetts: Is it out-of-date information? Is it duplicate or conflicting pages? Is it gaps nobody spots until a user asks? Or is it too many requests and no clear priorities?
26
00:05:50.680 --> 00:05:54.570
Toby Margetts: We'll give it about, kind of, 30 seconds to a minute for people to answer.
27
00:05:54.920 --> 00:06:01.029
Toby Margetts: And then we'll see, we'll see what comes up. What's, Jamie, what's your, what's your money on, if you had to, if you were a gambling woman?
28
00:06:01.230 --> 00:06:15.950
Jamie Sharp: Aha, which I am. It's really interesting, because we see all of these all the time, like, I'm actually genuinely really interested to see which one is the most common, because out in the field talking to people all the time, I see these come up. Every single one of these come up, actually.
29
00:06:15.950 --> 00:06:25.919
Toby Margetts: Yeah, 100%, totally agree. I think if I was pressed, I'd probably go, like, duplicate or conflicting pages. We do see a lot of that. I feel like that's something that's actually really hard to…
30
00:06:26.030 --> 00:06:31.549
Toby Margetts: Stay on top of, particularly when we've got, yeah, lots of devolved content teams as well.
31
00:06:34.510 --> 00:06:40.329
Toby Margetts: Right, we'll just give it a few more seconds, and then we will see, see where we're at.
32
00:06:44.790 --> 00:06:51.970
Toby Margetts: Nice one. Okay, that's interesting. There's quite a big split, which is, I think, particularly interesting.
33
00:06:52.030 --> 00:07:08.979
Toby Margetts: We've got out-of-date information is coming out, on top, 34%, duplicate or conflicting pages, 24%, gaps nobody spots until a user asks, 16%, too many requests, and no clear priorities at 26%, percent.
34
00:07:09.500 --> 00:07:24.760
Toby Margetts: I think this is really interesting, because ultimately, like, whichever answer actually wins, it creates a problem for AI search, as well as for content teams. So, if we think about out-of-date content, that really weakens freshness, obviously.
35
00:07:24.760 --> 00:07:31.750
Toby Margetts: Duplicate and conflicting pages weaken authority. Gaps make it harder to actually answer the question for AI.
36
00:07:31.840 --> 00:07:45.860
Toby Margetts: too many requests without priorities, that creates more content debt. So, the content problems that content editors are firefighting are often the same problems that actually stop that healthy content being found and trusted by AI.
37
00:07:46.090 --> 00:08:00.340
Toby Margetts: But yeah, really interesting to see such a broad split across the board, key point being that, yeah, all of these really contribute towards you needing to get your content in really great health, to ensure that AI can generate really great answers from it.
38
00:08:00.450 --> 00:08:10.579
Toby Margetts: I'm going to pass to Jamie now, who is going to talk to us a little bit about some of the discovery signals that AI uses to generate good answers. So, yeah, Jamie, back to you.
39
00:08:11.180 --> 00:08:13.739
Jamie Sharp: Thank you. So…
40
00:08:14.800 --> 00:08:33.039
Jamie Sharp: Content health and AI discovery, they're two separate topics. If the answer's buried, then AI effectively is going to struggle to extract it. If your page is poorly labeled, again, AI is going to struggle to extract that. And if you've got five pages that are saying slightly different things, then you're going to find, of course, AI's got less reason to trust any of those sources.
41
00:08:33.039 --> 00:08:46.319
Jamie Sharp: If your information is stale, then it's likely AI is going to choose another source. So, in our previous webinar, we called these the four AI discovery signals, so structure, metadata, authority, freshness.
42
00:08:46.460 --> 00:09:09.120
Jamie Sharp: And I hear a lot in everyday conversations with customers that this can feel a bit complicated, the language is quite technical, so if you're thinking that, don't worry, you're absolutely not alone. But it's actually really straightforward, it's really practical, and it's actually a lot less technical than, like, an SEO checklist, those types of things that we're more used to. So today, we're going to go through really actionable steps that you can implement tomorrow, and the important takeaway here is that
43
00:09:09.120 --> 00:09:26.989
Jamie Sharp: content health matters for AI discovery. It really protects those signals that help AI find, understand, trust your content. So we're going to focus on how teams keep those signals healthy. As Toby was saying, when the estate keeps changing, that's when it's really difficult. So we're going to jump onto the next slide, which we are talking about, the loop.
44
00:09:26.990 --> 00:09:30.850
Jamie Sharp: And walk through how you build those signals into day-to-day content work.
45
00:09:31.210 --> 00:09:44.250
Jamie Sharp: So this loop that you can see on screen, this is a practical operating rhythm that your teams can use and keep applying this checklist as your content changes. So the four signals here tell us what AI-ready content needs, which
46
00:09:44.250 --> 00:09:48.010
Jamie Sharp: As discussed is structure, metadata, authority, and freshness.
47
00:09:48.010 --> 00:10:12.849
Jamie Sharp: But keeping those things true after you've published, that matters just as much. So, sticking with our theme of giving you actionable steps to use with your team, and not just a load of buzzwords, we're going to use this loop to enable teams to decide what to do before creating new content, what to check before publishing, what to monitor afterwards, and then how to turn gaps into clear actions. So, just briefly touching on all of the steps in this loop, we've got plans, so you need to audit before you just
48
00:10:12.850 --> 00:10:18.960
Jamie Sharp: fall into the trap of adding more content. Publish, so obviously make the answer easy to find.
49
00:10:18.960 --> 00:10:35.100
Jamie Sharp: easy to understand and trustworthy. Monitor, so once you've got that content out there, then continuing to look for signs that the signals there are weakening, and then continue to improve. So turn those gaps into decisions. So if we jump to the next one, we'll jump into each of those stages in a bit more detail.
50
00:10:35.440 --> 00:10:54.219
Jamie Sharp: So this is the plan stage. So, planning really protects that structure, your authority, your freshness, by making sure your team answers the right questions in the right places, simply. So instead of just adding more content, and this phase really is about, like I said before, falling into that trap of adding more content. We see folks all the time thinking.
51
00:10:54.220 --> 00:10:55.770
Jamie Sharp: Okay, we know that…
52
00:10:55.830 --> 00:11:10.970
Jamie Sharp: an LLM or an AI model is looking for a particular answer, we can see our user base, our customers are struggling to find a certain answer, so we're just going to create more content to fill that gap, but actually adding to that content that can actually weaken your AI discoverability.
53
00:11:11.240 --> 00:11:18.280
Jamie Sharp: So before creating a new page, the most important thing to do is audit what you've already got. So check whether the answer already exists.
54
00:11:18.410 --> 00:11:31.400
Jamie Sharp: You need to agree which page owns it, and then you need to decide, do I need to create something new? Do I need to update what we've already got, or can I consolidate? And the important thing here is less duplication means a much clearer source of truth.
55
00:11:31.840 --> 00:11:33.859
Jamie Sharp: Let me jump to the next one, please, Toby?
56
00:11:33.990 --> 00:11:48.559
Jamie Sharp: So then this is Publish. So publish is not only about getting your content live, it's that moment where you make the page either easier or harder for the AI to read, for it to categorize, for it to trust, and importantly, you want it to choose your content.
57
00:11:49.240 --> 00:12:04.860
Jamie Sharp: So this is where it's really easy, and we see a lot of customers, a lot of folks making mistakes. So if the answer's buried, if the labels are vague, if the source of truth is unclear, if no one knows when it should be reviewed, that page is already weakening before it's even had a chance to perform.
58
00:12:05.860 --> 00:12:18.760
Jamie Sharp: So if we just talk through, again, those four key things, we've got structure, so you need to think about, here, is the answer clear? Is it specific? Is it near the top? And clear writing doesn't have to mean bland writing, but it's important to think about those steps.
59
00:12:19.190 --> 00:12:24.450
Jamie Sharp: Metadata, so if you've got titles, are the descriptions, labels, is the content type clear?
60
00:12:24.670 --> 00:12:40.620
Jamie Sharp: For authority, is this the source of truth? Is it aligned with related pages, or have you got slightly different conflicting information across these different spots? And the page needs to make the main answer explicit. Again, you don't want to be hiding the answer across multiple sections.
61
00:12:40.770 --> 00:12:52.650
Jamie Sharp: And then freshness. So, is there an owner? Do you have a review trigger? What's the update expectation on this? And a last updated date by itself doesn't necessarily mean that the content's fresh.
62
00:12:53.570 --> 00:13:04.219
Jamie Sharp: If you think about all of that, this is where technology really helps, so these checks are easy if you've just got one page, but if you're trying to do this across a whole estate, tooling can really help limit that
63
00:13:04.380 --> 00:13:22.159
Jamie Sharp: cognitive load when all of our teams are trying to do so much more all the time. It can really help you by flagging those vague labels, jumping straight into identifying where have you got missing answers, where's there unclear ownership at publishing time, and just take that mental load out of where to start and how to keep the content healthy.
64
00:13:23.090 --> 00:13:25.129
Jamie Sharp: If you jump into the next one, please, Toby?
65
00:13:25.280 --> 00:13:42.189
Jamie Sharp: So now we're looking at signals. So, the work doesn't stop at publications, you've done your audit, you've got your content in a good spot, and you've published it out, but a page can be really clear today. You feel really good about what you've got out there at this moment, but then it could become buried under a load of related pages next week.
66
00:13:42.450 --> 00:13:51.669
Jamie Sharp: You can be really confident, you've done your review, it's authoritative today, but then actually, a couple of months down the track, it needs to compete with 3 new campaign pages that have been released.
67
00:13:51.880 --> 00:14:11.799
Jamie Sharp: Again, you can be really confident that it's not out of date, and that everything is in a really good spot with regard to freshness, but then your figures update, policy changes, and again, your content's not in a great spot. So monitoring is all about looking for these practical signs that those signals are weakening, and it's all about focusing on that, handing your marketing content teams
68
00:14:11.800 --> 00:14:19.060
Jamie Sharp: Signals that they can recognise, so that this doesn't become another thing that they've got to be thinking about and worrying about as part of their already hectic days.
69
00:14:19.320 --> 00:14:36.820
Jamie Sharp: So again, we're looking for AI discovery signals, your prioritization inputs to help you decide where to focus. So the types of things you want to watch out for are important questions that are not answered clearly, related pages that are giving different answers, new pages that are appearing on an already existing topic.
70
00:14:37.130 --> 00:14:40.579
Jamie Sharp: Obviously, broken links, stale dates, outdated facts.
71
00:14:40.940 --> 00:14:50.770
Jamie Sharp: Important pages that might have unclear titles, headings, descriptions, search items, or analytics or feedback that show that people aren't finding those answers.
72
00:14:50.950 --> 00:15:01.740
Jamie Sharp: And then demand is how you decide what to look at first. So you've probably got a heap of data there, but then you're looking for possible evidence to action items. So some examples of that could be.
73
00:15:01.750 --> 00:15:20.280
Jamie Sharp: If you've got lots of missing answers, then that probably points to you, do you need to create something to fill the gap? If you've got lots of competing or conflicting pages, then that probably points to the fact that you need to look at that estate of what you've got and consolidate to make sure the answers are really clear, you've got a single source of truth. If you've got lots of outdated
74
00:15:20.350 --> 00:15:27.540
Jamie Sharp: Pages, then you need to update, and if you've got redundant or risky pages, then at that point you want to be looking at retiring some of that content.
75
00:15:27.760 --> 00:15:33.259
Jamie Sharp: And again, this is where technology really helps, because this is usually the first stage that we see where
76
00:15:33.290 --> 00:15:52.969
Jamie Sharp: where people break down, so if you've got hundreds, tens of thousands of pages, monitoring that by hand is a huge job, and you really do need technology for it, and there are tools to do this. We happen to make one, and it's built on real-life learnings that we've made, and from what we can see happening within the market with our customers globally, so in a little bit, we'll run you through that.
77
00:15:52.970 --> 00:16:04.929
Jamie Sharp: But the main takeaway on this point in the loop is that monitoring is just about spotting where the structure, the metadata, the authority, the freshness might be slipping before the estate becomes harder for people and for AI to use.
78
00:16:05.960 --> 00:16:09.199
Jamie Sharp: And then on to the final piece of the loop.
79
00:16:09.260 --> 00:16:31.830
Jamie Sharp: So this is where we're looking to turn gaps into decisions. This stage is where the loop helps teams choose the right content action, rather than, again, falling into that tempting trap of, oh, let's just create some more content. And the more you get into the routine and using the loop, the more it becomes second nature. So, things to think about is, if an important question's got no clear answer, you might need to create or update.
80
00:16:31.830 --> 00:16:36.349
Jamie Sharp: If the right page exists, but the headings, the titles are vague, you need to update that.
81
00:16:36.440 --> 00:16:46.449
Jamie Sharp: If several pages compete, or they contradict, you consolidate. If the content's stale, or it's critical, you need to update, you need to redirect, you need to retire. So.
82
00:16:46.450 --> 00:17:09.319
Jamie Sharp: the critical point here is not about reducing more content by default, it's about strengthening the content that you've got and the conditions that help AI find your content, understand it, trust it, and crucially, you want it to cite your content, the right content. So the more you get into that routine, you should see it as just a natural way of reducing unnecessary production, not adding to it. And again.
83
00:17:09.319 --> 00:17:17.449
Jamie Sharp: that's where technology really helps. So the hard part is deciding what to fix first, and that prioritization by topic, importance, page value, risk.
84
00:17:17.460 --> 00:17:20.459
Jamie Sharp: That's where getting a tooling can save the most time.
85
00:17:21.810 --> 00:17:35.370
Jamie Sharp: All right, I think we're going to jump into another poll at this point, so we are really keen to hear when you last audited your content. So we've got a few options that are going to come up, and again, keen for everybody to provide some answers if they can, so we're interested to see
86
00:17:35.370 --> 00:17:41.890
Jamie Sharp: Was it in the last 3 months? In the last 6 to 12 months? More than a year ago? Never or not that I know of.
87
00:17:42.060 --> 00:17:49.489
Jamie Sharp: So we kind of see most teams audit rarely, because it's a bit of a beast to do, which is why your content health can slip.
88
00:17:49.690 --> 00:18:04.509
Jamie Sharp: But really keen to see what everybody's thinking, so if, if you wouldn't mind just clicking the button while you get a sec. Toby, you're… you're really close to teams and talking about this sort of stuff all the time. Any thoughts on what you think the most likely winner will be out of these?
89
00:18:04.830 --> 00:18:12.569
Toby Margetts: I mean, yeah, I think they were probably, like, a long time ago, maybe more than a year ago, I feel like…
90
00:18:12.590 --> 00:18:26.509
Toby Margetts: content audit… I've never met anyone who, like, loves, gets really excited about a content audit. They tend to be, they tend to take a long time. They tend to be things that people put off, even though they are, or traditionally have been super important.
91
00:18:27.080 --> 00:18:34.219
Toby Margetts: So yeah, it's going to be interesting to see. I can totally understand why people do put them off, but yeah, we shall see.
92
00:18:34.390 --> 00:18:40.379
Jamie Sharp: I don't know, everyone's favourite job. We'll give it a couple more seconds, and then we will get
93
00:18:40.690 --> 00:18:42.200
Jamie Sharp: Get some answers up.
94
00:18:42.870 --> 00:18:56.699
Jamie Sharp: Alright, interesting. It's a bit of a split again, so I would say in the last 3 months is probably the highest, so we've got just over a third of folks saying in the last 3 months, which is really positive, because this is what everybody should be thinking, and it's really important.
95
00:18:56.740 --> 00:19:03.870
Jamie Sharp: In the last 6 to 12 months, it's just very close after that, and then we've got about a quarter of people saying more than a year ago, and about…
96
00:19:04.220 --> 00:19:28.300
Jamie Sharp: just over 10% of folks saying never or not that I know of. So, I would say, you know, if you're answering more than a year ago, that's really normal. As Toby said, like, audits are big, they take a long time, we see a lot of folks, they sort of become set and forget, they're, like, stale as soon as they finish. So, again, like, just getting into the habit of that loop, rather than feeling you need to do a big audit, is exactly where you need to be thinking. Small, regular, prioritized, let tech do the work for you.
97
00:19:29.750 --> 00:19:39.469
Toby Margetts: Yeah, absolutely. Yeah, really interesting, again, to see, to see a bit of a split in there. I think for those that, did say it's not happened for quite a while.
98
00:19:39.670 --> 00:19:44.350
Toby Margetts: I don't think that's a discipline problem,
99
00:19:44.500 --> 00:19:56.999
Toby Margetts: Yeah, it's a scale problem, I think, more than anything. It just becomes a really daunting thing to want to do. As Jamie mentioned, when you've got tens of thousands, maybe hundreds of thousands of pages, that's a super daunting thing.
100
00:19:57.050 --> 00:20:15.249
Toby Margetts: But everything I think that we've just talked about, is kind of quite straightforward on one page, as I said, but most teams, you know, are managing hundreds, thousands of pages, tens of thousands of pages across lots of different teams. There are campaigns, there are PDFs, there are service areas, and there's old content.
101
00:20:15.320 --> 00:20:29.529
Toby Margetts: And of course, it's not static, right? The estate keeps changing all the while that you're auditing it. That's why often content audits fail sometimes. You know, by the time you've started it and finished it, a bunch of content on your site's changed in that time.
102
00:20:29.580 --> 00:20:35.770
Toby Margetts: And that's why they often get put off, and why the ones that you do finish tend to go stale very quickly.
103
00:20:36.440 --> 00:20:49.660
Toby Margetts: Squiz Content Intelligence, which we've, talked a little bit about, helps by giving teams a really kind of current view of the estate, so surfacing the issues, prioritizing actually what it is that matters, and tracking that progress.
104
00:20:49.710 --> 00:21:00.199
Toby Margetts: as content changes. So, content health then actually becomes something you keep an eye on, not something you sort of schedule a yearly project on to actually go and have a look at.
105
00:21:01.910 --> 00:21:14.189
Toby Margetts: What I'm gonna do is jump into a quick video, which shows content intelligence in action. It'll show content intelligence, it will show a little bit about how it works with conversational search.
106
00:21:14.210 --> 00:21:22.430
Toby Margetts: It is a video, and I… apologies to everybody, I'm going to be doing the voiceover, so I'm going to probably be, clicking around a little bit and stopping.
107
00:21:22.530 --> 00:21:26.029
Toby Margetts: But I'll do my best to make it as seamless as possible.
108
00:21:27.280 --> 00:21:51.420
Toby Margetts: What we can see straight off the bat here is a view of the DXP console. Content Health is where content intelligence exists, so we're going to show somebody kind of clicking in there. We're actually using Squiz's website as the example here. We're getting a bit of an insight into, yeah, the AI readiness of Squiz's website. Not a particularly big website, probably quite small compared to a lot of people that are on the webinar.
109
00:21:51.540 --> 00:21:56.630
Toby Margetts: 212 pages, but nonetheless, it'll be interesting to kind of see what's in there.
110
00:21:57.370 --> 00:22:11.290
Toby Margetts: When we click in, we can see, kind of, straight off the bat when the last scan was done, when the next scan is due. You'll notice as well, straight off the bat, it's split into two different areas. So, on the left-hand side, we have accessibility, which we've not talked about loads yet.
111
00:22:11.700 --> 00:22:23.989
Toby Margetts: And then AI readiness on the right-hand side. I'm actually just going to jump into the accessibility side of things to begin with, and then I'll jump into the AI readiness. But it's good just to kind of see how holistic the tool is.
112
00:22:23.990 --> 00:22:31.709
Toby Margetts: A lot of people use tools for accessibility. Accessibility is super important. The content intelligence covers that in its entirety.
113
00:22:31.720 --> 00:22:45.529
Toby Margetts: We can see when we click into the accessibility auditor, straight off the bat, we get some really useful information, so we get a view of what your overall accessibility score is. Luckily, ours is excellent, would have been maybe a different story if it wasn't.
114
00:22:45.530 --> 00:22:58.690
Toby Margetts: But you can see what your score is, you can see how many issues that you've resolved since you last scanned. You can also see, you know, the percentage of pages that you have that have issues. Again, luckily, we're 99.1% of pages.
115
00:22:58.870 --> 00:23:14.529
Toby Margetts: don't have significant issues, which is obviously great. What we can do is we can actually filter by fix type as well, so you can see here, you can choose either code fixes or content fixes. In this particular example, we're going to have a look at some of the code fixes we might want to do across the site.
116
00:23:14.890 --> 00:23:27.070
Toby Margetts: under the kind of quick wins section here, this is showing us what things could we do right now that are going to have the most, kind of, profound impact on our accessibility score. So we could click into those and just, kind of, do those straight off the bat.
117
00:23:27.340 --> 00:23:33.750
Toby Margetts: As we scroll down, there are a bunch of, kind of, useful things related to why we're failing, where we're failing.
118
00:23:33.760 --> 00:23:51.549
Toby Margetts: We can also see, basically, all of the issues that we have across the site down here, and what a lot of people like to do is to filter by impact, so we can select the critical and serious issues here. We can see that there are three issues, really, that need our attention, and we can then click into one of those to find out a little bit more.
119
00:23:51.710 --> 00:24:10.599
Toby Margetts: Crucially, what this tool will do, it won't just tell you you've got issues, it will actually give you a solution to that issue, which I think is a bit of a differentiator. That's one thing being told, hey, you've got accessibility issues, and most people would go, yeah, I know, thanks. What do I do? How do I fix them, and in what priority do I need to, do I need to attack them?
120
00:24:10.600 --> 00:24:25.329
Toby Margetts: So this will tell you why there is a failure, it'll tell you the impact that's having on users, and crucially here, it gives you the fix, right? So there is the current implementation, here is the AI-generated recommended fix. We can copy that, and we can paste it straight into our CMS.
121
00:24:25.630 --> 00:24:28.500
Toby Margetts: And we get the fix, which is, which is great.
122
00:24:28.900 --> 00:24:48.859
Toby Margetts: So that's accessibility. Appreciate I've kind of rattled through that in quick time. Very happy if people want to talk to me in more detail about it and have a kind of deeper walkthrough of the platform from an accessibility point of view. Absolutely love to do that. What I'm going to show you now is the AI readiness side of things. So again, panel on the right-hand side when we go into Content Health.
123
00:24:49.680 --> 00:25:04.949
Toby Margetts: We're gonna click in, and we're going to see some really interesting information. So, the 9 out of 12 score on the left-hand side, that is essentially saying that content intelligence has determined that we have 12 sections of our site, or 12 topics, we call them.
124
00:25:04.950 --> 00:25:15.010
Toby Margetts: It's saying that 9 of my 12 topics are AI-ready, which is great. There are 3, however, that aren't AI-ready, and that we need to do something, something about.
125
00:25:15.970 --> 00:25:28.880
Toby Margetts: what it will do, as we scroll down, we can see what all of those topics are. Obviously, most of the people, watching this webinar aren't DXP providers. Your topics will likely be very different to, to ours.
126
00:25:28.880 --> 00:25:36.490
Toby Margetts: But what we're going to do here is just click into one, more or less kind of at random, to show you what happens when we click into a particular topic.
127
00:25:36.490 --> 00:25:41.099
Toby Margetts: And then where we can see the problems that we've got, and how we might be able to deal with those.
128
00:25:41.140 --> 00:25:56.420
Toby Margetts: So what I'm going to do here, we're going to click into the conversational AI searches a little bit better, because we're sort of talking about that as part of the webinar as well. Conversational AI search is a section on our site where we talk about it and what it can do for people, how it can improve user experience.
129
00:25:56.550 --> 00:26:02.659
Toby Margetts: And yeah, I'll show you a little bit about what it's telling us about that particular section of the site.
130
00:26:04.490 --> 00:26:19.710
Toby Margetts: So, in here, there are a few really interesting things. So, it is telling us, at the top here, for those who've got, kind of, eagle eyesight, there are 31 pages on our site which cover this particular topic, or have content related to this particular topic.
131
00:26:19.910 --> 00:26:36.429
Toby Margetts: It also mentions 2,771 questions. So it's really important to understand broadly how this tool works. What it's doing in the background is it is generating, 50,000 plus, sometimes hundreds of thousands of questions.
132
00:26:36.430 --> 00:26:51.479
Toby Margetts: related to the content on your site, and more broadly about the, sort of, the organization that you are. It is then asking those questions to the LLM, and it is generating answers, based off of the content that you have.
133
00:26:51.630 --> 00:27:11.330
Toby Margetts: where it is struggling to give answers to those questions, that essentially is where it flags that you've got issues, and it will explain, this is the reason that you've got issues. So, it's basically a really super advanced sort of Q&A tool where it says, here are 100,000 questions that somebody that could come into your website is likely to ask.
134
00:27:11.450 --> 00:27:22.260
Toby Margetts: Of that 100,000, you know, 99% were getting really great answers, 1% were not. In order to get that 1% up, we need to do, the following kind of, the following kind of things.
135
00:27:22.900 --> 00:27:39.550
Toby Margetts: So that's just a bit of an insight into broadly, how the tool is working. Again, what this will do, so within the conversational AI search section, it's telling us that there are 31 pages. We could click into that if we wanted to, and kind of assess it in a bit more detail at a page-by-page level.
136
00:27:39.570 --> 00:27:55.930
Toby Margetts: What I find most of our customers like to do, though, is to actually focus on where we've got priority issues here. So, regardless of what page they're on, this is telling us these are the issues that we recommend you go and sort out first, because they will have the biggest impact on your AI visibility score.
137
00:27:56.030 --> 00:27:58.720
Toby Margetts: We can obviously click View All and see all of them.
138
00:27:58.940 --> 00:28:12.760
Toby Margetts: We can also see all of the questions in this particular section, so just under 3,000. We can scroll through. I'm not going to sit here and make you look at 3,000 questions, but you can go in here as a content editor and get an understanding of the type of question that's being asked.
139
00:28:12.770 --> 00:28:26.439
Toby Margetts: see the response that is being given. I can search if I want to find something very specific that I'm looking for, but it allows us to kind of make that mental bridge between the section of the site and the type of questions that the tool determines are really important.
140
00:28:27.070 --> 00:28:33.909
Toby Margetts: What we can do is actually click on a particular example that it's called out. This is one of the really high priority, issues.
141
00:28:33.970 --> 00:28:49.689
Toby Margetts: And again, very much like the accessibility auditor, this doesn't just tell us that we've got an issue, it gives us, a suggestion. So, here it talks specifically about a problem, and this is very much, I've been told I'm not allowed to say, eating our own dog food, it's drinking our own champagne.
142
00:28:49.690 --> 00:29:01.149
Toby Margetts: It thinks that we have an issue related to this, so we should add a concrete example of answering a complex user query with the postgraduate scholarship scenario. It thinks we've got a gap there.
143
00:29:01.150 --> 00:29:21.719
Toby Margetts: It explains why that's an issue, and then it actually gives us a really explicit suggested revision. So it says, we think in this section of the site, you should add this copy. This will then allow the LLM to take that information, create a question and answer pair from it, and we'll then be able to deliver a really good answer to the customer.
144
00:29:21.920 --> 00:29:36.969
Toby Margetts: So, at a high level, it's doing this across your entire estate, which is why it's so good at scale, right? So, a user could go in and kind of manually do this, but across 10,000 pages, that becomes basically a full-time job, probably more than a full-time job.
145
00:29:37.280 --> 00:29:46.679
Toby Margetts: This tool scans everything, and then orders everything for you, tells you where you need to focus, and it just makes things much, much easier for the content editor.
146
00:29:46.850 --> 00:29:57.409
Toby Margetts: Finally, what I'm going to do is just show you the kind of final part of the loop, I guess. I'm just going to kind of click into a page on our website and show you conversational search in action.
147
00:29:57.410 --> 00:30:22.080
Toby Margetts: It's important to realize that the way conversational search broadly works on the front end is it uses that enormous database of Q&A pairings that the tool has generated to generate its answers from. And because they're all sort of compliant and verified, and we know that they're good, it means that the speed of the answer is really fast. We can be really confident that what's in there is correct, because content intelligence has assessed it all and said, yep, great, we can get answers to these
148
00:30:22.080 --> 00:30:29.620
Toby Margetts: questions. That makes the job of the LLM on the front end to provide answers so much quicker and so much more accurate as well.
149
00:30:30.090 --> 00:30:45.029
Toby Margetts: So we have it implemented just kind of in our, normal search. A user can go to the site, they can search, so how can Squiz help website engagement and conversions? We hit enter like we would any kind of normal search.
150
00:30:45.570 --> 00:30:57.900
Toby Margetts: And we get an AI answer to our question, and you can see it only takes a couple of seconds to get that answer. The answer streams, which is pretty cool. It tells us what sources it's generated that particular, that particular answer from.
151
00:30:58.010 --> 00:31:15.930
Toby Margetts: And as I said, we can be super confident, that the answers that it's given are accurate, because it's all based on content intelligence having verified them. We can, of course, ask follow-up questions, so tell me more about conversational search. I get an answer to my question.
152
00:31:16.020 --> 00:31:31.189
Toby Margetts: The way that we've implemented on the Squiz site, we have included the search results below as well. We call it, like, a hybrid implementation. You can see search results still exist down here. They don't have to, you don't have to have it implemented that way, it could just be a four-page kind of takeover.
153
00:31:31.190 --> 00:31:39.069
Toby Margetts: But a lot of people are somewhat wedded to their search results, I would say, and who are we to say that you should absolutely get rid of them straight off the bat?
154
00:31:40.360 --> 00:31:51.699
Toby Margetts: Alrighty, so that is a walkthrough of content intelligence, conversational search, hopefully a bit of understanding around how practically it can help you across a really, really broad, broad estate.
155
00:31:52.420 --> 00:32:10.579
Toby Margetts: What we've got at the moment is the ability to try Content Intelligence for a month. If that looks at all interesting, please do scan the QR code or get in touch with myself or Jamie directly. We would be happy to work with you to get that set up, if it's something that you're interested in.
156
00:32:10.580 --> 00:32:26.369
Toby Margetts: We do also, offer, sort of, the ability to have a bit of a look at some of your content, so we use some of the tooling to say, hey, we'll look at maybe a small size of your site to give you a bit of an insight into how you're scoring. Maybe that's a bit of a toe in the water.
157
00:32:26.370 --> 00:32:38.269
Toby Margetts: If you're not ready to sort of commit fully to content intelligence at the moment, although we do strongly encourage that customers do, this is going to be something that is so, so important, I think, for anybody who's got any kind of…
158
00:32:38.580 --> 00:32:42.059
Toby Margetts: Any amount of, content going forward.
159
00:32:43.270 --> 00:32:57.110
Toby Margetts: In terms of key takeaways, and what we've kind of run through today, there are sort of four key things. We talked up front about the fact that AI search is really kind of magnifying the existing content health problems that people are having.
160
00:32:57.390 --> 00:33:07.199
Toby Margetts: AI depends on having a strong structure, metadata, authority, and freshness. Jamie touched on that at length. It's so, so important to get those four aspects right.
161
00:33:07.310 --> 00:33:24.199
Toby Margetts: a repeatable loop that keeps those signals strong. We can't just treat these as, like, one-hit projects and go, yeah, great, we did that 3 months ago, we're gold. It's something that has to happen on a continual basis, because content is being changed and updated so frequently on people's websites.
162
00:33:24.200 --> 00:33:39.229
Toby Margetts: And yeah, finally, Squiz Content Intelligence is hopefully that tool that can really, really help you do that, to help you stay on top of it, to proactively scan your site, and to say, here are where you've got issues, this is how you go and deal with them.
163
00:33:39.230 --> 00:33:50.299
Toby Margetts: That's so useful when you've got content being constantly added to a site, potentially across devolved teams. Yeah, that, to our mind, beats doing content audits any day of the week.
164
00:33:50.520 --> 00:33:57.869
Toby Margetts: Given the nature of them, where they take maybe 3 months to run, and by the time you finish them, half the content on your site maybe has changed anyway.
165
00:33:59.710 --> 00:34:13.209
Toby Margetts: That's it from Jamie and I. I did see a flurry of questions flying in as we were going, which is amazing, and maybe we can jump into some of those and talk a little bit about them.
166
00:34:13.210 --> 00:34:21.860
Toby Margetts: I'm happy to jump in, Jamie. I don't know, Jamie, if you've got your eye on any in particular that you fancied answering. I did see one crop up, I think it was maybe the first one that came in.
167
00:34:22.020 --> 00:34:38.310
Toby Margetts: which was, in terms of freshness, can this be as simple as going into a page and updating slash republishing it? Will an LLM know that that's happened? Basically, yes, it can absolutely be as simple as that, so content intelligence would
168
00:34:38.360 --> 00:34:51.679
Toby Margetts: go in, and it would say, yep, you have got some out-of-date content here, we recommend that you go in and make it up-to-date. Of course, the beauty of that, using that tool is it will find all the instances of that.
169
00:34:51.710 --> 00:35:08.390
Toby Margetts: And then, yeah, the next time that you scan your site, Content Intelligence will say, yep, great, we don't have that issue anymore, the content is fresh now, it's not out of date, and it will proactively kind of keep you, keep you on track in that way. So yeah, in short, you're kind of, you answered your own question there, and LLM will.
170
00:35:08.390 --> 00:35:15.869
Toby Margetts: Once it re-indexes the page, it will be using the new information, because that's the most current. It won't be using, kind of, archived content.
171
00:35:16.720 --> 00:35:40.619
Jamie Sharp: Yeah, and I would maybe just add to that, you know, like, you don't want to think of freshness as just being performative, like, the readers and the governance processes could stop trusting it. So again, this is where, like, don't overthink it, stick to having that, like, methodical process, like we've demonstrated with the loop. Having a tool in there that can just direct you to where you need to make changes is the best thing to do, rather than just feeling like you need to make performative changes just to tick a box.
172
00:35:41.080 --> 00:35:43.570
Toby Margetts: Yeah, 100%, great, great shout.
173
00:35:43.670 --> 00:36:06.240
Toby Margetts: Yeah, to pull out maybe a couple more, what is the best way to decide which pages need attention first? It's a great question, and actually, like, so much of the, sort of smarts and intelligence that went into building content intelligence was around that kind of exact question. It was like, hey, there are sort of tools out there that are sort of okay at telling you what you need to do.
174
00:36:06.240 --> 00:36:14.959
Toby Margetts: But there aren't many that actually are really good at telling you where you need to start, and that's so crucial when you're dealing with, you know, a big website with thousands and thousands of pages.
175
00:36:14.960 --> 00:36:27.369
Toby Margetts: It can be overwhelming, even for content intelligence, to say, hey, yeah, you've got, you know, like, 50 critical issues, and you go, right, okay, where do I start? The tool will specifically call out the most profound impact to your AI readiness.
176
00:36:27.400 --> 00:36:44.030
Toby Margetts: will be, by changing these particular things. One thing that's actually really interesting is, so that shouldn't… there is a bit of nuance here, that shouldn't necessarily just be dictated by, like, having a bad score. So if there's a web page that has a particularly bad score.
177
00:36:44.030 --> 00:36:47.829
Toby Margetts: That isn't necessarily the first place to start, so…
178
00:36:47.870 --> 00:36:56.739
Toby Margetts: To give, like, a bit of an analogy, if that's not a very high-trafficked page, say 20 people visit that page a year, and the content's in bad shape.
179
00:36:56.740 --> 00:37:10.080
Toby Margetts: that's maybe not as big a priority as some content that's in better shape, albeit still not great shape, but that gets thousands and thousands of visits a year. Obviously, there's a bit of balance to be had there between how sort of visible and active that page is.
180
00:37:10.220 --> 00:37:15.919
Toby Margetts: And the tool will factor that in. It's kind of based on risk as much as anything, so…
181
00:37:16.030 --> 00:37:19.550
Toby Margetts: Not just about what's got the lowest score, it's about what's got
182
00:37:19.760 --> 00:37:24.869
Toby Margetts: The most kind of visibility and is having, you know, the most impact on your particular users.
183
00:37:25.340 --> 00:37:36.309
Jamie Sharp: Yeah, nice. We did get a question about a lot of our important information lives in PDFs, how should we fit those into a content health routine? And that question comes up a hell of a lot.
184
00:37:36.440 --> 00:37:46.939
Jamie Sharp: So, you know, like, as part of… as part of the process, we can absolutely scan those PDFs and provide answers on that, but what… what our strong advice would be is to try and keep
185
00:37:46.940 --> 00:38:09.379
Jamie Sharp: PDFs when it's required. So, you know, if you need formal documents, printable forms, signed publications, things like that, but don't try and force users or AI into extracting basic guidance from PDFs. Our strong advice with regard to PDFs, and I know it's difficult, especially for a lot of organizations that are sort of deeply embedded into having heaps of stuff in PDFs, but try and track PDFs as
186
00:38:09.380 --> 00:38:28.909
Jamie Sharp: assets with an owner and a review cadence, and prioritize trying to convert as many of your, sort of, frequently asked questions into proper content. That will make it way easier for you to be confident that you're getting the right answers, that it'll be really visible and well-cited by AI, and obviously making it much easier for human
187
00:38:28.910 --> 00:38:43.039
Jamie Sharp: users to also find the content that you want them to get at, so I think, you know, as part of this process, it's an important and really useful step to be thinking about, actually, like, what information should live in a PDF, and how best to handle that as well. Did you want to add anything to that, Toby?
188
00:38:43.360 --> 00:39:02.729
Toby Margetts: I don't think so, I think you covered it really nicely. Yeah, we, like, PDFs are a bit of a sort of dirty word, aren't they, these days? Ideally, you don't want to have them on your estate, but we understand that people do, and, you know, it's not easy to just say, hey, put everything into HTML. But yeah, I think outside of that, you covered it really well.
189
00:39:04.440 --> 00:39:15.609
Toby Margetts: one of the final questions in here is around, getting content intelligence set up, or how easy is it to get it set up? The answer, the short answer, I guess, is, like, it's very, very easy to set up, so…
190
00:39:15.610 --> 00:39:31.139
Toby Margetts: There's maybe, like, a sort of 3-4 step process, where you literally insert the URL of your site, and it will basically scan everything and all of the child pages underneath that, so most people would put their homepage URL in, and it will scan absolutely everything.
191
00:39:31.140 --> 00:39:46.040
Toby Margetts: You can choose to exclude things if you want to. Often people don't, because they want to see a view of everything, but sometimes there is a need to, maybe if there's a part of the site which is, I don't know, going through a ton of changes at the time, or whatever.
192
00:39:46.040 --> 00:39:53.819
Toby Margetts: often a few different reasons why certain things might want to be excluded. You can do that. You simply just enter the URLs of what you want excluded.
193
00:39:53.820 --> 00:40:18.340
Toby Margetts: You then have, like, a drop-down list where you say how frequently do you want to run the scan, so it could be quarterly, it could be monthly, it could be weekly. You could do it every day if you wanted to, if you had a real, kind of, sort of fast turnaround of how content is created. But yes, as I say, it's like a three, four-step process. You run your first scan, it will tell you all of the different topics that your site is comprised of.
194
00:40:18.390 --> 00:40:31.220
Toby Margetts: You then decide, of those topics, which ones you want to monitor in more detail. You might want to just look at everything, or you might say, nope, we want to focus on this particular part of the site to begin with, and work through it in whichever way that you see fit.
195
00:40:31.390 --> 00:40:40.489
Toby Margetts: But yeah, basically, it's super easy to get set up. We also have, a fabulous customer success team that onboard you in the tool.
196
00:40:40.490 --> 00:40:56.190
Toby Margetts: To be brutally honest, like, the tool is so easy to use, you don't really need customer success to onboard you, like, it is super straightforward and intuitive, but we do like to have customer success kind of just there when we get everything set up to make sure people feel comfortable and we have any kind of questions answered.
197
00:40:56.190 --> 00:41:05.619
Toby Margetts: But yeah, 99 times out of 100, I would say, customers are perfectly capable of, kind of, almost onboarding themselves, because the tool is so straightforward and intuitive to use.
198
00:41:05.830 --> 00:41:07.740
Jamie Sharp: Yeah, 100%.
199
00:41:07.740 --> 00:41:32.659
Jamie Sharp: I can see a couple more questions that popped into the chat, so I can see, and David's asked, can content intelligence be used for non-squiz matrix sites, such as WordPress? Tom's just jumped into the chat to say, yes, it can, but it absolutely can, so it's, completely tech agnostic, and obviously, again, we have lots of customers that use our full DXP suite, but then also, you know, oftentimes there'll be a swathe of, I think, content that they want to capture that's elsewhere, so yes, you can add
200
00:41:32.660 --> 00:41:38.010
Jamie Sharp: Absolutely plug it in, run it across your entire estate, doesn't matter what tech it's on, it'll do the same job across everything.
201
00:41:39.440 --> 00:41:48.900
Toby Margetts: Love that. I think, that's everything in terms of what came through in the Q&A. I'm not sure if there was anything else, Jamie, in the chat that we missed,
202
00:41:49.030 --> 00:41:50.540
Toby Margetts: If not…
203
00:41:50.770 --> 00:41:59.480
Toby Margetts: we can probably, probably call it there. I would just finish by saying, yep, thank you very much for, all of the engagement and the questions that have come through, that's, that's great.
204
00:41:59.530 --> 00:42:18.529
Toby Margetts: If you're interested in the tooling at all, or just have any kind of further questions about it, feel free to scan the QR code. Feel free to fire a message through to myself or to Jamie. Always happy to answer anything via email or jump on a call, do a demo to whoever would like to see it. Yeah, we're always on hand to help out.
205
00:42:18.670 --> 00:42:26.209
Toby Margetts: So yeah, Jamie, thanks very much for joining us today. It's been a great session, and yeah, we'll see everybody at the next one.
206
00:42:26.430 --> 00:42:29.759
Jamie Sharp: Yeah, look forward to seeing you all soon. Thank you, everyone. Thanks, Toby.
207
00:42:29.760 --> 00:42:30.850
Toby Margetts: Thanks, bye-bye.
Video: Watch the webinar (UK). Captions and transcript available on playback.
Poll Results

- Out-of-date information – 34%
- Too many requests and no clear priorities – 26%
- Duplicate or conflicting pages – 24%
- Gaps nobody spots until a user asks – 16%
- In the last three months – 33%
- In the last six to twelve months – 27%
- More than a year ago – 25%
- Never, or not that I know of – 15%
Watch the US & Canada webinar
Transcript: Watch the webinar (US)
1
00:00:29.540 --> 00:00:41.889
Megan Andrews: Hello, everyone! Welcome to our webinar today. I'm Megan Andrews, Director of Client Partnerships here at Squiz, and I'm joined by my colleague, Fran Zablocki, our Client Strategy Director.
2
00:00:41.890 --> 00:00:56.739
Megan Andrews: And today, we're gonna have some exciting content for you. Everything in this webinar is about being found by AI. So, really good tips to keep your content healthy, and just keep up with the pace of all the change we see today with AI.
3
00:00:56.920 --> 00:01:00.280
Megan Andrews: So, couple housekeeping items before we jump in.
4
00:01:00.510 --> 00:01:10.600
Megan Andrews: First of all, feel free to scan this QR code. You can access a one-month trial of Squiz Content Intelligence. So, get a little insight into your content, get a little head start.
5
00:01:10.600 --> 00:01:24.429
Megan Andrews: The other thing is, we will be recording this, so if you have to leave halfway through, or you want to share this out with colleagues after, you can do that. We will send out a link to the recording, and also any follow-up questions that we weren't able to get to today. So…
6
00:01:24.480 --> 00:01:28.180
Megan Andrews: Let's get into it, Fran. Exciting stuff today.
7
00:01:28.640 --> 00:01:35.409
Megan Andrews: Really quick introduction. If you're not familiar with Squiz, you probably are, but we're a digital experience company.
8
00:01:35.500 --> 00:01:49.989
Megan Andrews: So, if you're operating on any sort of scale with your organization, you're probably using lots of different tools across your tech stack, and the great part is Squiz has all these different types of tools to help you manage your digital experience across your site.
9
00:01:50.020 --> 00:01:57.220
Megan Andrews: You can find out more on our website, but certainly we'll talk a little bit more about one of those tools today.
10
00:01:58.630 --> 00:02:18.289
Megan Andrews: So, get out your pens, or your typewriters, and get learning today. We're gonna have four different areas that we really address. One is just, why does good content practice break down? If you're on a content team, you know this can happen, it's okay, it happens to the best of us. But we'll look at some of the reasons for that.
11
00:02:18.430 --> 00:02:24.349
Megan Andrews: And if you've joined our webinar series, the last one we gave was about these four AI discovery signals. So…
12
00:02:24.500 --> 00:02:32.029
Megan Andrews: Again, new information, a new era. We'll revisit those and talk through how that impacts your content and its discoverability.
13
00:02:32.140 --> 00:02:51.750
Megan Andrews: We will also look at how to use those signals in… in practice. What does that mean for how you're looking at your content, optimizing it, and keeping up to date with that content on your site? And then the fourth area is just how Squiz Content Intelligence as a tool can help you do this, and, you know, optimize and monitor that content.
14
00:02:51.750 --> 00:03:03.540
Megan Andrews: not just once, but at scale and repeatedly, right? So that's the part that's pretty hard, is content can be in great shape for a moment in time, but how do you continuously monitor it at scale?
15
00:03:05.710 --> 00:03:22.979
Megan Andrews: So, in theory, yeah, I think about this sometimes, Fran, where it's like these beautiful moments where maybe your team has done this website redesign, and you're so excited about it, you're like, ugh, our site's down to a thousand pages, and we know each one of those pages is thoughtful and curated and good to go.
16
00:03:23.040 --> 00:03:31.800
Megan Andrews: And for, I don't know, 6 months, maybe even a year, you feel pretty good about that, right? But over time, things start to happen.
17
00:03:31.830 --> 00:03:43.399
Megan Andrews: And, it's really hard to, over time, keep up with things, but also with just new best practices and things we're learning with this changing environment of large language models. So…
18
00:03:43.450 --> 00:03:54.810
Megan Andrews: Things we're used to, right, on content teams. You've got those beautiful pages, and then, oh, there's a whole new campaign! So we've got to build out this little, little sub-area of the site, new content, new requests there from different teams.
19
00:03:54.810 --> 00:04:04.409
Megan Andrews: The aging pages, look, most of us aren't in brand new companies or organizations. We've been around for a minute. Sometimes things have been, published for a decade or more.
20
00:04:04.410 --> 00:04:11.309
Megan Andrews: And it kind of stacks over time, too. You can imagine, different deadlines that come up across year to year.
21
00:04:11.310 --> 00:04:22.149
Megan Andrews: Maybe you have an orientation page that was… August 19th was an orientation day last year, but it's gonna be August 20th this year. Do both those pages still exist? Do they have conflicting content?
22
00:04:22.230 --> 00:04:26.429
Megan Andrews: And maybe your team has gotten really good at auditing.
23
00:04:26.630 --> 00:04:36.519
Megan Andrews: Once. And then, did you do it again? How stale was that audit after time? Did you have a good way to keep up with your auditing process?
24
00:04:36.740 --> 00:04:55.610
Megan Andrews: And I think all of us understand that, you know, the firehose of requests that can come your way, and just trying to understand, you know, where do we need to prioritize our time and our team's time to keep up with content. So, these are a lot of the common pitfalls, right, Fran? And I'm sure you've seen some of these across time as well.
25
00:04:55.610 --> 00:05:12.809
Fran Zablocki: Yeah, absolutely. You're reminding me of another thing that feels very similar, and that is gardening. Getting your whole garden nice and weeded and cleaned feels terrific, but it's always gonna need it again in, like, a week or two. So, good constant maintenance is key.
26
00:05:13.610 --> 00:05:24.940
Megan Andrews: It's true, and you want to… you know, like, finding the times, too, at scale where you need to bring other people in, recruit your daughter to come in and help you out on some of those days.
27
00:05:25.800 --> 00:05:26.880
Megan Andrews: Am I right?
28
00:05:28.310 --> 00:05:44.899
Megan Andrews: So, yeah, I think the other thing, too, is that, there… in a world of SEO, where we've been, there are ways around, content that's not great on your site. When I think about the impact of large language models and AI,
29
00:05:45.000 --> 00:05:49.969
Megan Andrews: Really, it's putting… casting a magnifying glass across every page on your site.
30
00:05:50.010 --> 00:06:02.329
Megan Andrews: And what I visualize is that moment in 2022, or whenever all these models were going out across the entire internet. They were indexing every single part of the internet.
31
00:06:02.330 --> 00:06:15.280
Megan Andrews: And so, if you had pages that existed and were published on the web, those were ingested by a large language model. Now, they're not constantly doing that all the time, but some are, and in different ways, and so…
32
00:06:15.320 --> 00:06:21.719
Megan Andrews: The real question is, when any of these new platforms, TrachiPT, Gemini, we're all using these.
33
00:06:21.870 --> 00:06:30.980
Megan Andrews: When users go to these sites and they ask a question, there's a lot of different paths in terms of what's gonna come back and be surfaced in that response.
34
00:06:31.120 --> 00:06:40.030
Megan Andrews: Now, some of these, models are relying on indexed information, or some combination of indexing and then going out and looking in real time.
35
00:06:40.330 --> 00:06:56.429
Megan Andrews: But what is true is that regardless of what's, being asked, once those models go out, if they can't find consistent, current, structured, clear information on a site, whether they did it, you know, 2 years ago, or they're doing it in real time.
36
00:06:56.530 --> 00:07:02.930
Megan Andrews: it's gonna skip over, that actual content. It's just not gonna pick it up, not gonna surface it in the response.
37
00:07:03.280 --> 00:07:16.979
Megan Andrews: So, if there's this stale, duplicate, again, things we talked about, conflicting information, it's outdated, you know, it's just a bypass. It's gonna go find some other source that's more relevant and better organized.
38
00:07:17.100 --> 00:07:28.529
Megan Andrews: So, while you might think you can get away with some things, those audience questions, they're gonna rely on good content underlying it, to give a clear, specific response.
39
00:07:31.490 --> 00:07:40.299
Megan Andrews: So, we'll engage with you all, because I am very curious. Which of these do you see most often in your own content?
40
00:07:41.010 --> 00:07:46.360
Megan Andrews: So we've got… is it out-of-date information? Duplicate or conflicting pages?
41
00:07:46.510 --> 00:07:49.080
Megan Andrews: Gaps nobody spots until user asks.
42
00:07:49.320 --> 00:07:55.410
Megan Andrews: Or too many requests and no clear priorities. Fran, I'm curious, what do you think we're gonna see here?
43
00:07:55.830 --> 00:08:20.789
Fran Zablocki: I suspect a little bit of everything. I imagine that everyone probably experiences a little bit of this on some level, but the one that comes up in conversation most is that last one, too many requests and no clear priorities. I feel like that's less of a content thing and more of a governance thing. A lot of times, the folks managing content on the website just have so many people asking them to make changes all the time, and it's hard to predict when that's coming in, so it's definitely been top of mind for people.
44
00:08:20.790 --> 00:08:22.770
Fran Zablocki: I've had lots of conversations on that point.
45
00:08:23.310 --> 00:08:32.590
Megan Andrews: Yeah, great point. And I've seen some duplicate pages, too, like, a good way to understand what's duplicated and what's going on there.
46
00:08:32.830 --> 00:08:36.189
Megan Andrews: Let's see, do we have the results? We do. Okay.
47
00:08:37.299 --> 00:08:53.300
Megan Andrews: Gaps, nobody spots until a user asks, about 40% of you. That's really interesting. Yep, how do we get insight into what's happening on those platforms? And you're right, that too many requests, no clear priorities, and out-of-date information, both at about 30%, so…
48
00:08:53.600 --> 00:08:59.750
Megan Andrews: Interesting. Well, you are all in line with each other, just common challenges right now.
49
00:09:02.830 --> 00:09:09.430
Fran Zablocki: Right, so what are some of the specifics on these discovery signals that
50
00:09:09.600 --> 00:09:22.190
Fran Zablocki: will tip you off that AI might struggle with, right? So, if you attended our previous webinar, we went in a deep dive on structure, metadata, authority, and freshness as the four AI discovery signals.
51
00:09:22.360 --> 00:09:37.129
Fran Zablocki: And there's some scenarios in which, AI is just gonna have a hard time, right? So, if you don't have clearly structured information, and this is best practice that has gone back a while, right? If you don't have structured headings, structured content,
52
00:09:37.170 --> 00:09:55.370
Fran Zablocki: if you don't necessarily have a single topic with subtopics in an organized way, AI is going to struggle with that, just like SEO struggled with that in the past, and just like human beings struggle to read, right? If it's not, you know, if it's not scannable, if it's not obviously organized, AI is also going to struggle with it.
53
00:09:55.370 --> 00:09:58.939
Fran Zablocki: On the metadata side, we need to make sure that
54
00:09:58.940 --> 00:10:08.899
Fran Zablocki: AI can understand what the page is about. And this also overlaps a little bit with traditional SEO, because we're looking at things like page title and URL path and
55
00:10:08.900 --> 00:10:13.630
Fran Zablocki: in particular, the meta description. I know in my experience that
56
00:10:13.630 --> 00:10:32.660
Fran Zablocki: One of the most common gaps is that either there's no meta description for the page, or there's a default meta description that's being used across tons and tons of different pages, and in both of those cases, that's gonna kind of erode AI's trust as to exactly what this page is about, and also that this page is unique and has, you know, a particular purpose.
57
00:10:32.670 --> 00:10:34.930
Fran Zablocki: On the authority side.
58
00:10:35.010 --> 00:10:45.300
Fran Zablocki: This is actually authority of what is on the page, not necessarily your brand authority or equity that you've established, there.
59
00:10:45.730 --> 00:10:51.070
Fran Zablocki: This is… Really making sure… that…
60
00:10:51.090 --> 00:10:59.429
Fran Zablocki: has a dedicated page. Probably the closest equivalent in the SEO world would be, like, having a canonical page, right? This is the one page that…
61
00:10:59.430 --> 00:11:24.389
Fran Zablocki: is the source of truth on this particular topic. And where authority can start to erode, and where AI may struggle, is if you have multiple pages that either have identical content around that exact topic, or really, really similar content. And this happens easily over time, right? People want to put as much information to be helpful as possible, but over time, several different units or departments replicate each other, and then have slight variations. And that may not be something that a human would pick up.
62
00:11:24.390 --> 00:11:35.230
Fran Zablocki: on, because it's spread out across the site, but AI will pick up on it right away, and it will erode that trust. It will think to itself, I'm not really sure which of these three versions to believe, and so I'm just not going to include you in the results.
63
00:11:35.510 --> 00:11:45.579
Fran Zablocki: And then the last one, on the freshness side, is pretty self-evident, right? Like, we want to make sure… this is really what we're focused on today, is, like, making sure that we're checking to be sure our…
64
00:11:45.680 --> 00:12:00.269
Fran Zablocki: content is up-to-date and hasn't been, you know, hanging out for 10 years, like Megan mentioned, or to make sure that we don't have multiple versions of the same thing with multiple dates. This is a really common issue as…
65
00:12:00.420 --> 00:12:18.240
Fran Zablocki: websites grow. You have annual or maybe semi-annual reports that go out, or information that goes out, whether if it's in higher education, you have courses that are being updated every semester, or whether it's financial aid information that has tables associated with it that get updated every year.
66
00:12:18.240 --> 00:12:33.109
Fran Zablocki: if you don't clear out the old versions of those, they're gonna end up competing with the new versions, and they're gonna confuse the AI answer, and in a lot of cases, erode that trust as well. So, that's, you know, some really typical scenarios in which
67
00:12:33.110 --> 00:12:40.979
Fran Zablocki: These discovery signals, if we're not doing a good job with upkeep, can start to, make it hard to show up in those results.
68
00:12:42.550 --> 00:12:46.879
Fran Zablocki: Alright, so, what's our solution here? It is a simple four-step.
69
00:12:46.900 --> 00:12:53.389
Fran Zablocki: process, a content health loop for AI discovery. Fairly straightforward. The first step is to plan.
70
00:12:53.390 --> 00:13:18.329
Fran Zablocki: It's always good to have a plan at the beginning. The second step is to publish, and that includes making any changes, that's new pages or edits to existing pages. The third step is to monitor and make sure that we're keeping track of how those changes and new pages are performing, and then once we see how they are performing, the fourth step is to improve and iterate and rinse and repeat, right? This is an ongoing cycle, and it's something that can be applied either in the micro.
71
00:13:18.350 --> 00:13:35.019
Fran Zablocki: for an individual page, or in the macro for an entire site. But as we'll talk about a little bit more, when you start to scale this up for a site that has hundreds or maybe thousands of pages, that's when it gets really, really labor-intensive and really, really difficult. I'm sure that's no surprise to any of you. I mean, I think that…
72
00:13:35.200 --> 00:13:42.529
Fran Zablocki: In my experience working with content for large institutions over 15 to 20 years now, I think
73
00:13:42.730 --> 00:13:48.549
Fran Zablocki: The reality is, and this is to make everybody… Feel a little bit better.
74
00:13:48.650 --> 00:14:03.650
Fran Zablocki: The reality is that most institutions don't even get to 100% of their content because they just don't have the time to get to 100% of their content. So not only is the tool that we're going to talk about today going to allow you to
75
00:14:03.800 --> 00:14:13.489
Fran Zablocki: save time on the pages you are looking at, it's actually going to make it possible to get coverage across the entire website, in some cases for the first time, or at least, you know, much more
76
00:14:13.490 --> 00:14:23.830
Fran Zablocki: readily than, you know, like, a major, major content effort would be, to do it manually. So, that's the content health loop. Let's drill down a little bit into the different steps.
77
00:14:23.830 --> 00:14:27.850
Fran Zablocki: On the first step plan, it's always good to have a plan.
78
00:14:27.850 --> 00:14:50.850
Fran Zablocki: it's always good to know where you are before you point to where you need to go, and so the first thing to do here is to just get that audit and get that assessment. And you can see we've taken some screenshots of our content intelligence tool to the right. That's what content intelligence does a really good job of. It gives you that macro view of overall site health and readiness from the AI side, but it also drills down into specific topics and pages.
79
00:14:50.850 --> 00:14:56.910
Fran Zablocki: In terms of where they're at and what kinds of improvements that they have. So… Planning really makes sure
80
00:14:57.000 --> 00:15:11.369
Fran Zablocki: you just know what you have, and what it contains, and you can start to identify things, even at the planning stage, that are going to be easy fixes, like really old content, or duplicate content, or things that are missing, things like, you know, the correct meta description, like I mentioned earlier.
81
00:15:12.030 --> 00:15:21.010
Fran Zablocki: The next step is publish, and this is really where AI discovery signals can be strengthened, right? This is where the change is actually taking place.
82
00:15:21.010 --> 00:15:27.630
Fran Zablocki: I talked a little bit about each of those discovery signals and what, you know, can erode AI trust.
83
00:15:27.630 --> 00:15:42.580
Fran Zablocki: But this is where you're addressing things like structure and metadata and authority and freshness, and we're trying to make those answers easy to find and understand and trust. So this is really you getting into the page, into the CMS, and making the changes.
84
00:15:43.040 --> 00:15:47.620
Fran Zablocki: The next step is to monitor, and this is…
85
00:15:47.700 --> 00:16:00.619
Fran Zablocki: The thing that easily falls through the cracks, if it's something that we have to do manually, and why something like content intelligence is so valuable, because even after you've published, it's going to be taking a regular look at everything that changes.
86
00:16:00.620 --> 00:16:08.160
Fran Zablocki: And it's going to be re-ranking things to show you how much they've improved, both on the overall site level and on the page level.
87
00:16:09.080 --> 00:16:26.560
Fran Zablocki: I like to divide up the monitoring plan into prioritization of pages, right? So, like, you don't necessarily have to monitor 100% of your pages every week or two, but you probably want to monitor your really, really important pages.
88
00:16:26.560 --> 00:16:44.540
Fran Zablocki: every couple weeks or every month, and you want to get to everything at least once a year, and I'd even recommend, like, every 6 months. We've… there's been some studies out that have shown that, like, after 6 months, AI's signals for freshness really start to erode, so it's looking for things that have been published within the last 6 months. So,
89
00:16:44.610 --> 00:16:56.660
Fran Zablocki: Anyway, it's possible to actually look at certain pages more frequently. You might look at the most important ones 3 or 4 times, before you look at some of the least important ones once. Now.
90
00:16:56.660 --> 00:17:09.360
Fran Zablocki: how do you know which ones are most important and least important? One way is to take a look at your overall traffic, and analytics can help that. But I would take that traffic, and I would cross-reference it with what you know to be your core
91
00:17:09.359 --> 00:17:32.089
Fran Zablocki: customer journeys, or student journeys, depending on your industry. So, you know, let's take higher education, for example. A prospective student is going to need some really top-line information in order to enroll. They're probably going to need to look at all the admissions content, your academic program content, maybe your campus visit content. So start with all the different pages that belong on that path.
92
00:17:32.090 --> 00:17:46.469
Fran Zablocki: And make those a priority. And then make those pages that are getting a lot of traffic a priority, and, you know, when you cross-reference both, it's high traffic, and within that critical path, it should be the first group that you take care of, and so on down the line.
93
00:17:46.820 --> 00:17:52.900
Fran Zablocki: Alright, next step is… to improve. And so, once we know what has…
94
00:17:53.480 --> 00:18:04.409
Fran Zablocki: A gap, we want to make sure to decide what we need to do with that, and some common scenarios that you might run into here, and considerations for those scenarios.
95
00:18:05.860 --> 00:18:15.399
Fran Zablocki: R. Finding, an important question does not have gate, because it's going to say, here's… here are the questions that, you know, people are asking.
96
00:18:15.500 --> 00:18:24.460
Fran Zablocki: here's the questions that I, as an AI tool, am asking, that I'm getting, you know, I'm getting answers that aren't quite good enough.
97
00:18:24.460 --> 00:18:44.280
Fran Zablocki: 2, and so that will point out which pages need to create a focused answer, whether it just doesn't exist at all, and you might need a new section on a page, or a completely new page, or whether it's just that the content of the page just needs an update. Perhaps it's to change the H2 title
98
00:18:44.280 --> 00:18:54.610
Fran Zablocki: to the form of a question, and tailor that next paragraph that's after that H2 to more directly answer the question. That's kind of a practical step there.
99
00:18:54.610 --> 00:19:09.850
Megan Andrews: One thing I'd like to point out here and remind people of, too, because you make such a good point that generally as humans, right, we don't necessarily think in terms of… unless we're content people. Like, oh, that H2 could be different, or we could ask it, you know, phrase that a different way.
100
00:19:10.090 --> 00:19:21.929
Megan Andrews: this is… this tool is driven by a large language model giving an evaluation, so it's actually really helpful. We don't always, as humans, think in those terms. We might look at a page, visually assess it, right?
101
00:19:22.070 --> 00:19:24.639
Megan Andrews: And as a content team, think, we're good.
102
00:19:24.920 --> 00:19:39.099
Megan Andrews: But this is something we might miss, right? Where it's like, actually, the structure of this page, underlying this page needs to be addressed. I think that's helpful to keep in mind as you think about these things. There's two ways to design things, humans and bots, right?
103
00:19:40.070 --> 00:19:52.030
Fran Zablocki: Yeah, definitely. And they don't have to be at odds with each other, you just have to make sure that you're doing the right thing for both sides. And there's a lot of overlap, but there are some particulars around AI that, yeah, humans would not necessarily notice.
104
00:19:52.040 --> 00:20:00.229
Fran Zablocki: So, next scenario is several pages competing or contradicting. Kind of talked about this with, you know, source, ultimate source of truth.
105
00:20:00.230 --> 00:20:23.529
Fran Zablocki: what's the practical solution to that? Consolidating into one source of truth. That could be really explicit, like, you're just gonna make this page the only thing that has this content on it, but more than likely, you're gonna take one of the pages, make it the source of truth, and then look at the other pages that had repetitious content, and change that content to a reference to the source of truth page, right? So instead of trying to go into all the detail on that page.
106
00:20:23.530 --> 00:20:28.159
Fran Zablocki: You just want to create a crosslink to that page that mentions that particular topic.
107
00:20:28.260 --> 00:20:43.809
Fran Zablocki: And the last one is just content that's stale or expired or critical, and this is just that gardening analogy where we just need to make sure we're clearing out the old stuff, and we don't have old dates and figures and financials sitting around that are causing things to get
108
00:20:44.090 --> 00:20:58.940
Fran Zablocki: messed up in the, in the results. I typically take a rubric for this, that if anything is over a year old and has really low traffic, it is a prime candidate to retire.
109
00:20:58.970 --> 00:21:16.740
Fran Zablocki: there's gonna be exceptions. There's usually some really old pages that just have legs, because they're an article that somebody wrote that gets traffic annually, perennially, and, you know, you want to make sure to keep that. But use your analytics, use your, you know, age of page.
110
00:21:16.740 --> 00:21:27.960
Fran Zablocki: Use traffic to determine what can be retired, because tightening up the pages themselves and making sure that you don't have bloat is going to help with overall results.
111
00:21:28.550 --> 00:21:31.819
Fran Zablocki: Alright, so those are the four steps. Sorry, go ahead, Megan.
112
00:21:31.820 --> 00:21:45.429
Megan Andrews: You know, one of my favorite examples, we had talked about this a little earlier, was, I think… here's a practical question for you. So let's, many of you may know, like, Steve Jobs gave a very famous commencement speech, and…
113
00:21:45.550 --> 00:21:57.409
Megan Andrews: It was super popular, it's popular across YouTube, all these different places. It happened to be at Stanford University, and Stanford hosts a page that has that, commencement speech recording on it.
114
00:21:57.550 --> 00:21:59.999
Megan Andrews: and trans, transcribes it. So…
115
00:22:00.030 --> 00:22:03.929
Megan Andrews: It's a very popular page that people visit a lot. That…
116
00:22:03.950 --> 00:22:20.609
Megan Andrews: That speech was given in 2006, right? So, that's 20 years ago now. What do you do with that type of page? To your point, there are these types of pages, really popular articles, things like that. How do those sit with the freshness perspective? What do you do about those?
117
00:22:21.570 --> 00:22:32.100
Fran Zablocki: That's a really, really good question, because you can't really change the original publishing date, right? Like, you want to keep that, so it's got the historically accurate time.
118
00:22:32.100 --> 00:22:52.419
Fran Zablocki: But I could see actually creating some additional promotional content that is fresh and has newer dates to it to call attention to that… that page, right? And, like, referencing that page from a couple of different places. Or maybe, you know, create a new news article that is revisiting, you know, Steve Jobs
119
00:22:52.600 --> 00:23:07.419
Fran Zablocki: speech from 20 years ago in the light, you know, in today's context, or today's light, and then, you know, that kind of, like, surfaces it in a more modern context. If it's something that is, you know, an evergreen page and not a news article.
120
00:23:07.420 --> 00:23:19.659
Fran Zablocki: then just changing the content on the page and changing the dates will… that'll be it. But yeah, it's kind of a balance between, like, maintaining the historical record, but also, like, calling attention to it with some new, fresh content.
121
00:23:20.330 --> 00:23:21.040
Megan Andrews: Nice.
122
00:23:22.780 --> 00:23:28.220
Fran Zablocki: Alright, so, here's the honesty question. When did you last audit your content?
123
00:23:29.810 --> 00:23:30.339
Megan Andrews: Oh, boy.
124
00:23:30.340 --> 00:23:45.839
Fran Zablocki: be afraid to answer never, because that is actually a fairly common answer, right? Like, maybe you just haven't been able to get around to it, and that is exactly why we have tools like Content Intelligence to help you be able to get, you know, get through that thing that you just haven't had time to do.
125
00:23:46.740 --> 00:23:53.420
Megan Andrews: I like the never or not that I know of. You could claim… I mean, I'm not aware of an audit.
126
00:23:54.410 --> 00:24:05.449
Megan Andrews: But this is great. I will be curious, again, be honest in this, it's really good to understand what people are, are actually experiencing. And we do have a question, too, that I'd like to get to, Fran after this, related to…
127
00:24:05.620 --> 00:24:16.080
Megan Andrews: to content summaries. You know, there are lots of different ways and tools. There's everything from spreadsheets to something like content intelligence, but here we go.
128
00:24:16.440 --> 00:24:28.040
Megan Andrews: Oh, 20%, never, not that I know of. Totally fine. Most people sitting right around either in the last 3 months or 6 to 12 months, so that's pretty good. 30% each. Good job, everyone.
129
00:24:28.040 --> 00:24:30.330
Fran Zablocki: Pretty even split across everybody.
130
00:24:30.330 --> 00:24:31.060
Megan Andrews: Yeah.
131
00:24:31.300 --> 00:24:32.110
Fran Zablocki: Good work.
132
00:24:33.020 --> 00:24:47.809
Megan Andrews: Now, one question that came in that I think could be great to address here, and it's more a comment, but we could riff on this for a moment. We are a two-year community college, our SEO tool customer success manager mentioned AI overviews.
133
00:24:47.810 --> 00:24:54.679
Megan Andrews: And ways to get found in them. One of the ways is to have good quality content. So, AI-generated content is usually vague.
134
00:24:54.960 --> 00:25:00.709
Megan Andrews: And so this is a question around, or comment around AI-generated content.
135
00:25:00.940 --> 00:25:09.990
Megan Andrews: So, guess that's something to call out here. We're not necessarily saying, use this tool to generate content, right? It's very different from that.
136
00:25:10.650 --> 00:25:17.230
Fran Zablocki: Yes, yeah, I think… Using AI to generate content is only…
137
00:25:17.530 --> 00:25:26.319
Fran Zablocki: useful as a very first draft, and I think a lot of other people kind of share that, sentiment. Like, for example, the content intelligence tool is going to suggest
138
00:25:26.460 --> 00:25:30.219
Fran Zablocki: What you might… how you might want to position new content, but…
139
00:25:30.600 --> 00:25:50.299
Fran Zablocki: ultimately, you're the editor, right? Like, you're the one who needs to make sure that you agree with that statement, and that, you know, that's something that you want to publish, but I wouldn't be our recommendation to necessarily flood your site with AI-generated content. In fact, in a lot of cases, even the company… I was just reading an article yesterday, actually, that the companies…
140
00:25:50.300 --> 00:25:52.570
Fran Zablocki: Who are providing.
141
00:25:52.580 --> 00:26:05.670
Fran Zablocki: easily created AI content are now starting to realize just, how much has been flooding their platforms, right? Like, LinkedIn just put in a button that said, like, please report AI slop.
142
00:26:05.860 --> 00:26:29.740
Fran Zablocki: And so, I think it's really important for it to be human-created content. Maybe AI-assisted, but human-created, ultimately, in the end. And yeah, so we're, you know, like, our system is going to identify where things need to be improved. It is using AI to help you do that, but the actual content generation should definitely have, as they say, like, a human in the loop, and the human is the last step.
143
00:26:31.080 --> 00:26:31.740
Megan Andrews: Yep.
144
00:26:35.550 --> 00:26:39.320
Fran Zablocki: Alright, so, we know it's a lot of work.
145
00:26:39.450 --> 00:26:54.889
Fran Zablocki: We know it sometimes takes too much work to do, and we don't get to all of it, and that's why we created Content Intelligence. So, good practice at scale. We just want to be able to do all of these good practices that you've probably been doing manually, and create a tool
146
00:26:54.890 --> 00:27:05.500
Fran Zablocki: that can do this at scale across the entire site, and then maintain it across the site, so you just feel like you're on top of it, at all times, and not falling behind. And so…
147
00:27:05.500 --> 00:27:11.329
Fran Zablocki: On one page, very easy to do. On 10,000 pages, not so much. So…
148
00:27:11.520 --> 00:27:14.969
Fran Zablocki: Some of the other challenges, it's not just volume, it's also just…
149
00:27:15.380 --> 00:27:32.819
Fran Zablocki: how many people you have working on content, and where they're located, and how much you communicate. If you're in a larger institution that's decentralized, you might have 100, 200 different people managing content on the website. Coordination amongst them can be tricky. People will have different levels of
150
00:27:33.000 --> 00:27:56.480
Fran Zablocki: familiarity with all of these topics and content management in general, and so the level of quality and consistency, maybe with your branding and your voice and tone, might be, you know, there might be some gaps there. That's how we end up with multiple pages answering the same question. It's also how we end up with, you know, review dates and owners that are all over the place and that get really hard to track and centralize.
151
00:27:56.490 --> 00:28:01.080
Fran Zablocki: Publishing models may allow certain people to publish, but they might be…
152
00:28:01.130 --> 00:28:10.450
Fran Zablocki: are kind of breaking their rules or not practices. So overall, changes happen faster than manual audits can keep up, and once you actually have the audit.
153
00:28:10.700 --> 00:28:25.359
Fran Zablocki: in place, it's very difficult to take, you know, a thousand recommendations and decide which of those thousand to start with without some kind of prioritization list, and that's another key component of content intelligence that we'll… we'll show you shortly.
154
00:28:27.520 --> 00:28:29.450
Megan Andrews: Let's show them. How about that?
155
00:28:31.100 --> 00:28:41.520
Megan Andrews: We'll do a little video for you. Now, just a heads up, the video moves quickly, so our voices will as well, but you'll get a good idea of what exactly content intelligence looks like.
156
00:28:42.020 --> 00:28:50.110
Megan Andrews: So, if you were to log into your SquizDXP and go into the Content Health area, this is the backend, this is what you'd see currently.
157
00:28:50.180 --> 00:29:08.439
Megan Andrews: And this tool actually is split, right, into accessibility and AI readiness. So, it's really top of mind for everyone right now. Turns out good accessibility is critical to AI readiness and visibility. So, that's why we've paired them together. We'll take a look at the accessibility piece first.
158
00:29:08.690 --> 00:29:17.189
Megan Andrews: At the top, you're gonna get your score, your kind of roll-up summaries of everything that's going on across your domain, or the site that you've plugged in here.
159
00:29:17.190 --> 00:29:33.770
Megan Andrews: And then, I love this piece. Is it a content fix? Is it a code fix? Do I need a developer or a content editor? Really easy way to filter those quick wins, depending on who's in there. And then just snapshot of some different patterns and things it's seeing across your site.
160
00:29:33.960 --> 00:29:44.119
Megan Andrews: And of course, that big old scary long list, but don't worry, you can filter it by just focusing on what is the most critical thing we need to do, or most serious update.
161
00:29:44.120 --> 00:29:56.860
Megan Andrews: or violation here on the site. And clicking into those, it gives you all the context you could want, probably more than you would ever need, but you will deeply understand what exactly is going on, the impact it has.
162
00:29:56.860 --> 00:30:06.399
Megan Andrews: And it shows you, side by side, what's on your site right now, and what is that recommended fix that would, alleviate this issue and resolve it for you. So…
163
00:30:06.720 --> 00:30:10.700
Megan Andrews: That's a really helpful way to go through the accessibility side of your site.
164
00:30:10.860 --> 00:30:13.000
Megan Andrews: And Fran, how about AI?
165
00:30:13.380 --> 00:30:32.849
Fran Zablocki: Right? Alright, so accessibility is for humans, and the AI readiness is for the bots, is, like, what we like to say. Same kind of overview, right, giving you an overall health rating, then breaking it down into the individual topics and showing you how well you're performing. Now, this is showing…
166
00:30:34.050 --> 00:30:41.979
Fran Zablocki: Squiz.net is our website, so, priority issues, but we do have some remaining, and even without the high priority issues.
167
00:30:42.090 --> 00:31:06.820
Fran Zablocki: identified with an AI-ready section, there's plenty of detail contained within, right? So in this conversational AI search topic, there's 31 pages that are included in that topic, and we have 4 low-priority issues that we need to deal with. Each one of those is detailed in that list there, and you can drill down and see exactly what it is that's the issue. This next part, where it says 2,771 questions, is where it is emulating and showing exactly how that fan-out quiz
168
00:31:06.820 --> 00:31:08.439
Fran Zablocki: It's working, and so…
169
00:31:08.440 --> 00:31:20.320
Fran Zablocki: you can look through every individual question and see how it's being answered. It's really quite impressive. Now we're looking at an individual issue detail and some suggested revisions, so…
170
00:31:20.330 --> 00:31:40.229
Fran Zablocki: to that question earlier, it's just suggesting what you should include, and, you know, really, you should take that and edit it and publish it as you like. And it's also showing all the different pages that are referenced for this issue, and then if you, if you want, there's a link out to the specific page of your website to just make it easy to look at.
171
00:31:40.230 --> 00:31:42.580
Fran Zablocki: Now, we're looking at the Squiz
172
00:31:42.660 --> 00:31:58.189
Fran Zablocki: conversational search interface installed on our site. Feel free to go to Squiz.net and try it out for yourself. It's the best way to kind of get used to it. But this is the newest version of our search that incorporates this type of AI
173
00:31:58.210 --> 00:32:17.009
Fran Zablocki: conversational search functionality that we've gotten used to with things like ChatGPT right into the experience and flow of your website, right? So, it's got your answer in there, it's showing the different sources, it's also a place to do follow-up questions so that you can create a conversational record there.
174
00:32:17.050 --> 00:32:37.439
Fran Zablocki: And it's still paired with the traditional search results as well. So, it's very rich, kind of giving you the best of both worlds. You get that conversational AI response, but you also have that faceted search set that, you know, we've all been used to over the years, and between those two things, you're really getting a rich set of information for the questions that you're asking.
175
00:32:39.020 --> 00:32:41.020
Fran Zablocki: Okay, let's see where we go next here.
176
00:32:41.240 --> 00:32:42.729
Megan Andrews: Nice. Wow. Thank goodness.
177
00:32:42.730 --> 00:32:43.820
Fran Zablocki: Okay, here we go.
178
00:32:43.820 --> 00:32:45.250
Megan Andrews: I think that's it.
179
00:32:45.290 --> 00:33:01.100
Megan Andrews: No, we did it, and it's… you know, it's a lot of… it's a lot to take in. I think, probably a lot of this is just understanding how to equip your team, right? There… there's a manual approach that's gonna really be painful, and so…
180
00:33:01.100 --> 00:33:13.050
Megan Andrews: We're just talking through one possible solution and tool here. If you do decide you're excited and want to try Content Intelligence for a month, scan that QR code again. It will give you access to that,
181
00:33:13.080 --> 00:33:16.639
Megan Andrews: not only the AI search visibility, but that accessibility auditor.
182
00:33:16.720 --> 00:33:22.730
Megan Andrews: And give you those prioritization of fixes. And it works with whatever stack you're using, it will work.
183
00:33:22.900 --> 00:33:37.169
Megan Andrews: So that's the great piece. It is super easy to do, no matter where, where your content lives at this point. We do have a question that came in, will this video be shared afterwards? Yes, it will. So if you're registered for
184
00:33:37.170 --> 00:33:59.749
Megan Andrews: the webinar, that, email address you put on file, we're gonna go ahead and blast out everything that we did here today, including the recording. So, a look back. Key takeaways. What in the world did we talk about today? Really big point is just that AI, the AI world, it's really putting magnifying glass on your content, and that includes your content health issues or problems.
185
00:33:59.750 --> 00:34:10.539
Megan Andrews: Again, those signals that you really need to be aware of, because AI depends on them. It's strong structure, metadata, authority, and freshness. So hopefully we talk through some different good examples for you there.
186
00:34:10.639 --> 00:34:29.679
Megan Andrews: And then, a repeatable way to keep those signals strong, they could be great for a moment in time, and then over time, we've seen, right, different examples of how that could, decline. And Squiz Content Intelligence helps you do that, not just as a snapshot, but really at scale over time.
187
00:34:30.980 --> 00:34:36.230
Megan Andrews: So we'll see if we have any more questions. I saw a couple more come in in the chat.
188
00:34:36.380 --> 00:34:45.490
Megan Andrews: And let's see, here's one. Does having a page not indexed affect AI's ability to crawl that page? Fran, what are your thoughts there?
189
00:34:48.449 --> 00:34:49.210
Megan Andrews: Oh.
190
00:34:49.219 --> 00:34:58.259
Fran Zablocki: Sorry, I'm muted. I've been trying to find that article, which we're going to share momentarily. It was a New York Times article, so, so we'll follow up on that.
191
00:34:58.389 --> 00:35:06.999
Fran Zablocki: if you have something listed as noindex, it's not going to get picked up. So that is actually something that came up in our previous webinar as, like, the very first
192
00:35:07.229 --> 00:35:23.749
Fran Zablocki: gate post for being picked up by AI search results, and that is that if you've basically, you know, closed it off, as noindex, or if it's something that's behind a login, right? So if it's on an intranet or portal, it's just not gonna… not gonna show up. It's not gonna be able to get access to that.
193
00:35:24.560 --> 00:35:34.369
Megan Andrews: I had a great example of this, actually. We were putting the AI tool across one of our early higher education sites, a big university, and we were covering the topic of campus life.
194
00:35:34.370 --> 00:35:44.800
Megan Andrews: And we're trying to figure out all the URLs that were involved in that specific topic, and we thought we had them all. And pretty much towards the end of our auditing process, it was a little bit more manual a year ago.
195
00:35:44.800 --> 00:35:56.669
Megan Andrews: One of their content editor said, hey, wait, there's this whole page, it is an FAQ just about campus life and everything we'd been trying to look at, but it wasn't indexed anywhere. There was no breadcrumbs to get to it, it was just…
196
00:35:56.810 --> 00:36:10.960
Megan Andrews: floating out there in the ether. So we… you know, sometimes these processes, right, reveal where you do have these gaps. Hopefully just… you can speed up that time to where you know where the gaps are. I think that's the point of having a good tool.
197
00:36:12.260 --> 00:36:13.310
Megan Andrews: Alright.
198
00:36:14.000 --> 00:36:21.840
Megan Andrews: Do we have any other questions that came in? Let's see… The Q&A here.
199
00:36:22.010 --> 00:36:24.349
Megan Andrews: We'll do one more, I think we've got time.
200
00:36:24.810 --> 00:36:25.890
Megan Andrews: like that.
201
00:36:26.470 --> 00:36:29.520
Megan Andrews: So this came in earlier, Anne,
202
00:36:31.250 --> 00:36:39.270
Megan Andrews: How often should we review our most important content? Like, what's the… what's the pace at which we should be thinking about these audits and reviews, Fran?
203
00:36:41.640 --> 00:36:55.130
Fran Zablocki: Well, like I mentioned before, it really depends on how high priority the pages are, right? I think the range should be that the most important pages are potentially even every week, right? If you're running a newsroom, or you're running,
204
00:36:55.380 --> 00:37:11.729
Fran Zablocki: an events calendar, or you're in the midst of a really intensive period, like, if you're, for example, in higher education, you know, you're entering the fall, and, like, orientation is hitting, and people are signing up for classes, and there's just a ton of activity and a ton of people looking for, information.
205
00:37:11.950 --> 00:37:29.999
Fran Zablocki: It could be daily, but just for short periods of time, and just for really, really important pages. But that's not practical for most scenarios, and so that would be, like, a small subset of pages. So, maybe daily for that kind of scenario. Most likely every month for really important pages, you know, in normal times.
206
00:37:30.020 --> 00:37:49.599
Fran Zablocki: every quarter, maybe, for that secondary tier of priority. And then, I'm recommending every 6 months for the… for all pages on the site, and that is mainly because, you know, that freshness is such a huge key for it getting picked up by AI. You can do all… you can check all the other boxes and have all the great… the best information.
207
00:37:49.600 --> 00:38:00.009
Fran Zablocki: But if AI looks at two equally weighted pages, and one of them is, you know, a year or a year and a half old, and one of them is within, like, 3 months or 5 months of being published.
208
00:38:00.010 --> 00:38:08.460
Fran Zablocki: It's likely to pick the one that's more current, so it is important to be able to take a look at everything, at least within 6 months, a year at the very most.
209
00:38:09.540 --> 00:38:17.710
Megan Andrews: Nice, love that. Another one just came in, not sure if it was covered earlier, did you cover the importance of an FAQ schema markup in the context of AEO?
210
00:38:17.880 --> 00:38:22.929
Megan Andrews: what new words we're all working with here, AEO included?
211
00:38:23.290 --> 00:38:26.829
Megan Andrews: One quick thought I've had on this is that, I had one…
212
00:38:26.970 --> 00:38:42.800
Megan Andrews: client we were working with who was like, well, what if we just do all of our FAQs, they aren't visual for, you know, like, on the front-end display, it's just schema markup where, you know, large language models can learn what they need to learn and go on their way. So what's your thought on that, Fran?
213
00:38:43.300 --> 00:38:51.839
Fran Zablocki: Yeah, so schema markup is something interesting, and my… I mean, my simple way of understanding it in my own brain is…
214
00:38:52.030 --> 00:39:06.000
Fran Zablocki: what you provide in markup, in code, in schema, should reflect, not, like, near-identically, but it should reflect what is visible on the front end, in the design, and in the content that's being
215
00:39:06.000 --> 00:39:20.050
Fran Zablocki: read by human beings, right? So if you have a page that has a couple questions and answers, like maybe you're using an accordion component to do that, or maybe just the page itself has H2s that are in the form of questions with paragraphs that follow.
216
00:39:20.390 --> 00:39:29.149
Fran Zablocki: you can make the argument that you'd want to include the FAQ schema at least inline on that page. But what I have seen is kind of…
217
00:39:29.150 --> 00:39:41.659
Fran Zablocki: an overcorrection because of how many articles and SEO firms have been talking about how schema is really important, and it's sort of like the old days where people were, like, just putting keywords everywhere.
218
00:39:41.660 --> 00:39:50.359
Fran Zablocki: some of the recommendations of, like, let's just add all the schema and every single page and load it up, and that's overkill. And actually, Google has said that,
219
00:39:50.460 --> 00:40:02.369
Fran Zablocki: not that it's gonna penalize you necessarily, but it's definitely not helping your cause if you're just stuffing too much schema in there. So, you know, my practical approach would just be take a look at the page that you have right now.
220
00:40:02.370 --> 00:40:06.100
Fran Zablocki: If it looks good, if it's doing what it needs to do for humans who are on it.
221
00:40:06.100 --> 00:40:27.240
Fran Zablocki: ask yourself which schema makes sense to include, because in a lot of cases, there's some pretty low-hanging fruit, right? You might have that FAQ schema, or maybe for higher ed, if it's an academic program page, there is an academic program schema that should be included on every academic program page. If you have events, there's an event schema, and so on and so forth. So there's a layer of kind of like, okay, let's just match up the schema
222
00:40:27.240 --> 00:40:45.370
Fran Zablocki: to the things that we already have, and I would take that as step one, and then, you know, I think you can start to think about adding additional schema as a second pass. And if you've done something like Content Intelligence Review, and you've added more content to answer those questions, and things are being ranked as healthy.
223
00:40:45.590 --> 00:40:52.889
Fran Zablocki: that's the point at which you would want to go back in and say, okay, are there… is there any more tweaks we can make on the schema side?
224
00:40:52.890 --> 00:41:05.999
Fran Zablocki: So, it's kind of a layered approach, and I wouldn't necessarily just jump in and be like, oh my gosh, we have to add, you know, we have to add all the schema to all the pages just to try to move the needle, because it's one… it's only one signal out of many, many, and you can kind of actually overdo it.
225
00:41:06.370 --> 00:41:09.959
Megan Andrews: Yep. Yep. It's a good thing you keep in mind,
226
00:41:10.140 --> 00:41:19.600
Megan Andrews: And I think these best practices will continue to evolve, so it's good to experiment, to your point. Try some different things, make sure, measure them, see if they're working.
227
00:41:19.770 --> 00:41:28.990
Megan Andrews: One last question that came in, and then we'll wrap this up. How long does it typically take AI to drop search results for content that isn't easily accessible or outdated?
228
00:41:29.290 --> 00:41:33.519
Megan Andrews: I guess one interpretation of that may be also is, you know, what…
229
00:41:33.790 --> 00:41:40.759
Megan Andrews: Is, is, are these large language models just gonna completely ignore things that are not accessible or outdated?
230
00:41:42.700 --> 00:41:45.340
Megan Andrews: I think probably a lot of them would, yep.
231
00:41:45.340 --> 00:42:08.379
Fran Zablocki: I think so. I think, actually, there's a lot of parallels between how a screen reader interprets the content on the site and how AI crawls the site. And again, so this is a big overlap, and another reason why accessibility is such a core piece of this and getting that right. And I know a lot of you have already been doing a really good job with accessibility and really staying on top of it, and in that respect, you've already done a lot of the work you need to to be
232
00:42:08.380 --> 00:42:12.149
Fran Zablocki: To be on top of it, but if you do have accessibility issues.
233
00:42:12.150 --> 00:42:16.449
Fran Zablocki: you're kind of… you kind of have now a compound problem, right? Like, not only are you…
234
00:42:16.450 --> 00:42:39.070
Fran Zablocki: not accessible for screen readers, but AI is going to end up not being able to navigate as well, because a lot of the same structures and signals are being used by that. So, it's part of the reason why AI… sorry, why accessibility is one half of our review, and AI readiness is the other half, because you really have to have both of those buttoned up, in order to get the best results.
235
00:42:40.910 --> 00:42:48.800
Megan Andrews: Definitely. Well, if there are any other questions that you all come up with, we'll make sure we, answer them in our summary email to everybody, but
236
00:42:48.800 --> 00:43:01.730
Megan Andrews: Thanks for joining us! It's a lot of information to take in, and you know, even just having these conversations, starting to get the vocabulary and framework for this, and use this with your team, sometimes that's helpful. Watch it together, talk about these things.
237
00:43:01.730 --> 00:43:17.790
Megan Andrews: And get that, common language and understanding going with your team. But we do often have these webinars, they're very frequent. I think monthly we've got them, so keep an eye out for the next one. And we'll keep diving into these different topics, but thanks for joining us, and we'll see you next time, everyone.
238
00:43:17.790 --> 00:43:19.389
Fran Zablocki: Good to spend time with you, thanks.
Video: Watch the webinar (US). Captions and transcript available on playback.
Poll Results

- Gaps nobody spots until a user asks – 39%
- Out-of-date information – 28%
- Too many requests and no clear priorities – 28%
- Duplicate or conflicting pages – 6%
- In the last three months – 31%
- In the last six to twelve months – 31%
- More than a year ago – 19%
- Never, or not that I know of – 19%
Catch up on part 1 and register for part 3
How AI is changing the way clients discover expertise - and how to check whether your thought leadership is ready.
How to build AI readiness into the way your firm plans, creates, approves and activates thought leadership.