Being found by AI: Tips to keep your content healthy

1
00:01:17.880 --> 00:01:26.989
Joel Goodman: Good morning, everybody! Welcome to our webinar today, Being Found by AI Tips to Keep Your Content Healthy.

2
00:01:27.040 --> 00:01:39.340
Joel Goodman: My name is Joel Goodman, I'm the VP of Growth Strategy here at Squiz. I'm based in Nashville, Tennessee, so it's actually evening for me, but, really excited to be with all of you. With me is, my…

3
00:01:39.420 --> 00:01:48.969
Joel Goodman: Actually, one of my oldest friends in this business of web, but also now a colleague, Fran Zablocki, our client strategy director. Hi, Fran, how's it going?

4
00:01:49.740 --> 00:01:57.259
Fran Zablocki: It is going well, Joe. Good morning to all. I am in Rochester, New York, so it is also good evening to me, but glad to be here with all of you.

5
00:01:58.540 --> 00:02:13.330
Joel Goodman: Just a couple of quick housekeeping notes up front, if, well, you've registered for this webinar because you're here, you're gonna get the recording of this afterwards, so, don't worry about that. If you want to send it on to friends, we really appreciate it, like, please do that.

6
00:02:13.330 --> 00:02:31.580
Joel Goodman: Today's, we're just, we're just gonna cover some, tips to keeping your content healthy. It's meant to be very practical, we're not gonna go into broad AI theory or anything like that. This is really just how to keep the content across your websites healthy for AI visibility, for usability, for the people that visit your website.

7
00:02:31.580 --> 00:02:36.710
Joel Goodman: And just general maintainability for your own teams, whether they're small.

8
00:02:36.790 --> 00:02:53.340
Joel Goodman: or large. So, you know, most content teams already know what good content should do, it needs to be useful and accurate, it needs to be easy to find, easy to maintain, but we all recognize how hard it is to, keep up with requests, keep up with.

9
00:02:53.340 --> 00:03:07.550
Joel Goodman: how well that, that usefulness and accuracy kind of stays when multiple people are editing content or creating content, when your business priorities change. And that's what we kind of want to talk about with you today.

10
00:03:07.920 --> 00:03:13.880
Joel Goodman: Who is Squiz? We're a global digital experience company, hence why Fran and I are…

11
00:03:13.970 --> 00:03:33.689
Joel Goodman: here in the US, talking to you in Australia, New Zealand. We help enterprises build brilliant digital experiences fast on a platform that embraces change, and we are super excited to show you, some of, well, one, some of the thinking we've been doing around content, as well as how to, we've practically applied that to the tools that we're building.

12
00:03:35.460 --> 00:03:46.770
Joel Goodman: So today, a few things that we're gonna look at. One, why good content practice quietly breaks down those ways that kind of get into the good intentions that you have when you're planning content.

13
00:03:46.770 --> 00:04:02.389
Joel Goodman: We're going to look at four AI discovery signals, and then we want to look at how to use those signals in the content health loop to make sure that you're able to see how AI is being discovered, and then work that into your content to make it better and more discoverable.

14
00:04:02.390 --> 00:04:12.640
Joel Goodman: And then finally, we want to show you Squiz Content Intelligence, which is the tool that we've built to help you in that content loop, making all of that content manageable at scale.

15
00:04:12.880 --> 00:04:31.099
Joel Goodman: It's meant to, really show you those AI discovery signals. If you've come to a previous webinar, you'll probably have heard, some of that. We want to get into more practical uses for you, and really help you work this into your day-to-day workflows, for your organization.

16
00:04:33.170 --> 00:04:37.400
Joel Goodman: So, first I want to look at why good content practice breaks down.

17
00:04:37.520 --> 00:04:59.909
Joel Goodman: You already know this, mentioned this at the top, like, the problem is keeping all of your content accurate, and relevant and fresh all throughout the year. Good content practice breaks down when you have new requests that come in, you have unclear ownership or governance, you have pages that are aging, and you don't know which ones they are, and how often you need to go back and update that content.

18
00:04:59.910 --> 00:05:17.950
Joel Goodman: you might have stale audits, or duplicated content in different places across your website, or a lack of prioritization across all of this. You know, these are familiar principles, you know, your content should be clear, and it should be accurate, useful, current, findable, and owned.

19
00:05:19.070 --> 00:05:41.050
Joel Goodman: you might think of, like, a campaign page that repeats information that's on a product page. You can slowly find those pieces of information conflicting with each other. That can be difficult to spot and difficult to track, but it also makes it difficult for AI search to find and understand which of those points is accurate, which is correct, and which it should serve up in its answers.

20
00:05:41.050 --> 00:06:03.739
Joel Goodman: It could be also a PDF that remains live someplace after your web page has changed, and you may find that you're not able to keep those consistent, and so you have basically dead content that's living on, and where do you find it, you know? AI search, I mean, the truth of this is that AI search really puts a magnifying glass on all of these problems.

21
00:06:03.740 --> 00:06:05.180
Joel Goodman: If you…

22
00:06:05.180 --> 00:06:16.030
Joel Goodman: use AI in any capacity, you've probably seen this. You know, it amplifies the language that everyone uses. That's how we get generative AI. It also looks at all of the

23
00:06:16.030 --> 00:06:29.549
Joel Goodman: the problems and the discrepancies across our content, across the ways that we operate, and makes it even louder. So, for all the good that it actually does, there's a whole other side of it that just amplifies those areas that

24
00:06:29.550 --> 00:06:32.010
Joel Goodman: We need to focus on.

25
00:06:35.430 --> 00:06:52.599
Joel Goodman: And this is a little flowchart about how that works, right? So, you might have an audience question, and your healthy content is the one thing that needs to be consistent, needs to be current, needs to be structured, needs to be clear, so that that person gets a clear answer when they're searching on ChatGPT or on Google with AI mode turned on.

26
00:06:52.600 --> 00:07:02.670
Joel Goodman: You know, or being found by AI starts with healthier content, but that healthy content isn't something that teams can guarantee through, like, a one-off check here and there, right?

27
00:07:02.670 --> 00:07:08.089
Joel Goodman: Ai search doesn't remove the need for good content operations, but it does make them more important.

28
00:07:08.090 --> 00:07:20.970
Joel Goodman: If your content estate contains duplicate, stale, or conflicting answers, AI systems have a weaker foundation to work from, and this comes down to trust signals. So, if your content's stale, if it's duplicative, if it's conflicting, if it's outdated.

29
00:07:20.970 --> 00:07:28.010
Joel Goodman: the AI doesn't know how to trust that content. It doesn't know which points are real, which points are the right points, and so…

30
00:07:28.010 --> 00:07:41.989
Joel Goodman: the way that large language models work is when they get confused, they either make something up, and that's why you see hallucinations, or in the case of especially third-party search, if you're talking about your chat GPTs and your perplexities and your clods.

31
00:07:41.990 --> 00:07:51.269
Joel Goodman: they're gonna go somewhere else where they can trust that content, where those trust signals are stronger, and they're not getting confused. And that's… that's really what…

32
00:07:51.270 --> 00:07:54.710
Joel Goodman: Healthy content, works to prevent.

33
00:07:55.990 --> 00:07:58.220
Joel Goodman: So I'm gonna kick off with a poll.

34
00:07:58.900 --> 00:08:00.189
Joel Goodman: We love a poll.

35
00:08:00.490 --> 00:08:17.569
Joel Goodman: So which of these do you see most often in your content? We want to talk about A out-of-date information, B, duplicate or conflicting pages across your website, gaps that nobody spots until a user asks, or too many requests and no clear priorities.

36
00:08:17.580 --> 00:08:25.259
Joel Goodman: Fran, which ones are you seeing as you work with, with clients, across these different, content operation pieces?

37
00:08:26.730 --> 00:08:45.799
Fran Zablocki: You know, I'm always, interested in the governance side of things, and it's that too many requests and no clear priorities that really becomes the true challenge, I think. It's just… I mean, this is overwhelming sometimes in its scale and the amount of work that it is to keep a website up and all that content.

38
00:08:45.800 --> 00:08:58.260
Fran Zablocki: And adding in the layer of trying to find… be visible to AI can feel like an additional burden on top of something that's already really heavy. And so, you know, I feel… I think a lot of times…

39
00:08:58.270 --> 00:09:13.060
Fran Zablocki: it's the other processes and people that, at the institution or organization, who mean well, but are just kind of, you know, peppering people with lots and lots of requests. And it can just kind of feel overwhelming. So I've heard that a lot recently in conversations.

40
00:09:18.300 --> 00:09:27.429
Joel Goodman: Alright, we're looking at 42% have out-of-date information across their websites. We've got some of you with a mix of duplicate or conflicting pages and gaps.

41
00:09:27.600 --> 00:09:33.560
Joel Goodman: We're seeing a lot of these things, in and across all of the content that we…

42
00:09:33.730 --> 00:09:42.749
Joel Goodman: look at every day. And so, this is… this is, this is pretty par for the course. Like, people… people are seeing this across every…

43
00:09:42.870 --> 00:09:46.350
Joel Goodman: single… Sector that we work with.

44
00:09:51.850 --> 00:10:10.280
Fran Zablocki: Alright, so I want to talk a little bit about some of the different content health problems that are causing a weakening of AI discovery signals. If you have listened to some previous webinars, you'll remember that we have four major signals, structure, metadata, authority, and freshness, that

45
00:10:10.280 --> 00:10:19.740
Fran Zablocki: we evaluate content on, to make sure that it's ready for AI, and each of those kind of has fail points or weakening points. So, for structure.

46
00:10:20.480 --> 00:10:37.210
Fran Zablocki: the main point is to extract a clear answer, right? And so, with that, you're looking at structured headings, structured content, and making sure that each page is focused on only one topic with a number of different subtopics.

47
00:10:37.210 --> 00:10:54.260
Fran Zablocki: a page is having a tough time understanding… or, sorry, if AI is having a tough time understanding what the page is about, it's likely a metadata problem. And this is going to be familiar for anyone who has worked in SEO as well, where page titles, page URL structures, if you have meta descriptions that are either blank.

48
00:10:54.260 --> 00:11:10.009
Fran Zablocki: or duplicate, with a number of different pages, that can really trip up AI, just as it does humans and SEO. Authority is interesting, because if you have brand authority, and you have authority with the site, you've gotten a long way already.

49
00:11:10.010 --> 00:11:24.560
Fran Zablocki: However, authority in this case for AI is just how trustworthy the single source of truth is on a particular page, and it can get eroded by having duplicate content on more than one page, or content that's really, really similar.

50
00:11:24.560 --> 00:11:36.939
Fran Zablocki: on more than one page. So, for those of you who are familiar with, you know, canonicals in the SEO world, AI needs to be able to trust that you have a single source of truth for things, and if you have slightly different

51
00:11:37.090 --> 00:11:50.519
Fran Zablocki: versions of very similar information on different pages, that's gonna start to erode things, too. And then freshness, I think that was, you know, one of the ones a lot of people responded to in the poll. This one is pretty self-evident, right? Like, just making sure that you have

52
00:11:50.520 --> 00:12:05.819
Fran Zablocki: Current content, making sure that pages that have publishing dates have the most recent publishing dates, making sure, especially if you have reference information, like tables and figures and processes, that the latest versions of those are showing up, and kind of,

53
00:12:06.190 --> 00:12:24.520
Fran Zablocki: in that vein that older versions of documents and processes and procedures are cleaned out of the website, so they're not accidentally showing up there. AI really appreciates current content, and anything that's beyond, like, 6 months old starts to really weaken that signal.

54
00:12:27.700 --> 00:12:38.069
Fran Zablocki: So, today we want to talk a little bit about a content health loop for AI discovery, and this is a simple four-step process that can be used at scale, but it can also be used

55
00:12:38.070 --> 00:13:02.040
Fran Zablocki: at the micro level, so you could take this approach for an individual page, you could take this approach for sections of a particular website, or you could take it across the entire website. The first step being planning things out, then step two being publishing, then we want to monitor what we've published and see how it's performing against what we'd hope for, and then iterating and improving

56
00:13:02.040 --> 00:13:06.759
Fran Zablocki: And then repeating the cycle from there. I'll go into a little bit more detail on each of these.

57
00:13:08.680 --> 00:13:17.970
Fran Zablocki: The first step, plan, is really the most important to get started, and it can be a little bit tricky

58
00:13:17.970 --> 00:13:38.109
Fran Zablocki: to know where you need to go if you haven't already figured out where you are today, right? And so, having an audit of what you already have, having an honest review of what's working and what's not, is the first step. And this can be before creating a brand new page, or it can be when you want, you know, before doing some kind of modification on an existing page.

59
00:13:38.110 --> 00:13:42.719
Fran Zablocki: You can see to the right the content intelligence tool.

60
00:13:42.720 --> 00:14:03.940
Fran Zablocki: is going to help with that at scale by looking at all the pages on the site and starting to organize them around topics, and identifying where those individual topics are either performing well in terms of AI visibility or poorly. That kind of initial audit and assessment is really necessary in order to be successful with all the steps that

61
00:14:03.940 --> 00:14:04.840
Fran Zablocki: beyond.

62
00:14:07.560 --> 00:14:18.370
Fran Zablocki: The second step is publish, and this is really where most of the work and most of the impact on the ultimate result for people, for SEO, and for AI search.

63
00:14:18.370 --> 00:14:43.200
Fran Zablocki: is involved, and this is where we get into those AI discovery signals, and where we have the ability to strengthen them. So, whenever you're going to be publishing new information, we want to make sure that we have the appropriate structure and metadata, like I mentioned before, that you don't duplicate content from another page, and that you have this as a source of good information. If this isn't the page that is the authority on something, and it's referring to

64
00:14:43.200 --> 00:14:56.739
Fran Zablocki: content that's better owned by another page, it's fine to have reference content, and it's fine to link over to it, but we just want to make sure that we're not duplicating large amounts of content that are exactly the same from page to page, and of course, freshness.

65
00:15:00.870 --> 00:15:21.749
Fran Zablocki: We've published our page, or our section, and now we need to make sure that we're monitoring it over time to make… to see if we're successful with the changes that we've made. You can see to the right that we have, in the content intelligence tool, an accessibility and an AI readiness health check. It's gonna give you a top-level overview of overall health.

66
00:15:21.750 --> 00:15:43.360
Fran Zablocki: But it's also going to drill down into those different areas of accessibility rules and topics on the AI readiness side, and it's going to let you know where those things are either strong or weak. We'll go over a short demo of that in a little bit. But you really need to set up regular content health checks to keep your best content trusted, and that's easier said than done, right? We don't…

67
00:15:43.540 --> 00:15:58.279
Fran Zablocki: even doing the content health checks and setting up those schedules can feel like a task that can get overwhelming, and the way that I've usually handled that when I've been in charge of content is to prioritize the most important pages with the most,

68
00:15:58.280 --> 00:16:06.999
Fran Zablocki: or the least amount of distance between each review, right? So let's say you have an editorial calendar, you have a review calendar, and the top…

69
00:16:07.000 --> 00:16:24.090
Fran Zablocki: 15-20% of your website needs to be reviewed every 2 weeks or every month. Maybe the next 25 to 30% needs to be reviewed every quarter, and then the bottom 50% needs to be reviewed annually at a minimum. That is one way of managing

70
00:16:24.430 --> 00:16:31.060
Fran Zablocki: a really large web estate, and I usually… Organize that around…

71
00:16:31.180 --> 00:16:40.899
Fran Zablocki: How much traffic particular pages are getting, but you also want to look at it from the standpoint of which pages are part of your core

72
00:16:40.900 --> 00:16:50.049
Fran Zablocki: conversions and tasks that people are trying to do, and which ones are part of your key audience pathways. And also which pages may…

73
00:16:50.050 --> 00:17:05.640
Fran Zablocki: be performing really, really well, in terms of web traffic, even though they're kind of tucked away and buried within the web structure. So, it's a combination of using a tool like Content Intelligence and things like Google Analytics for overall traffic, to prioritize what

74
00:17:05.640 --> 00:17:12.520
Fran Zablocki: You know, what, pages you need to do from month to month and quarter to quarter, but ultimately.

75
00:17:12.780 --> 00:17:25.219
Fran Zablocki: Like we've been saying, it can be quite a lot of work to put that together, and it can be quite a lot of returns to the same pages over the course of the year, which is why we came up with a tool that can really help lift that heavy weight.

76
00:17:28.500 --> 00:17:34.639
Fran Zablocki: And then the last piece, after we've published our content and we've evaluated it and measured it against what we hoped.

77
00:17:34.720 --> 00:17:49.860
Fran Zablocki: were… are to turn any gaps that have been identified by the tool or by our own eyes into decision-making points. And this is the act of actually improving content page to page. So, you may find that you had an important question that

78
00:17:50.190 --> 00:17:51.989
Fran Zablocki: Didn't have a clear answer.

79
00:17:52.240 --> 00:18:04.799
Fran Zablocki: That means we need to create a focused answer or update an existing page. Perhaps we just need to tweak something, maybe the information is actually on the page, but we need to put it in the form of a question and an

80
00:18:04.800 --> 00:18:14.000
Fran Zablocki: answer. Maybe the header needs to be rewritten as a question, and the answer needs to be rewritten as the paragraph below it.

81
00:18:14.450 --> 00:18:19.730
Fran Zablocki: We may have a useful page, we know there's good content on it, we know that it's authoritative content.

82
00:18:19.730 --> 00:18:42.529
Fran Zablocki: But we're also noticing that it's not showing up as well, or performing as well as we thought. In that case, we want to take a look to see if any of the metadata is an issue. So, do we have structured headings? Do we have the appropriate page title, or another title that's showing up in the code? Do we have meta descriptions at all? And if we do have those meta descriptions, are they actually describing what is on…

83
00:18:43.140 --> 00:18:45.110
Fran Zablocki: page.

84
00:18:45.350 --> 00:18:56.369
Fran Zablocki: Again, like I've mentioned, we only have pages that are competing or contradicting. In that case, you may want to remove the contradictions altogether, or turn them into more direct references.

85
00:18:56.370 --> 00:19:06.160
Fran Zablocki: Or consolidate all of them into one page, and even remove the other pages entirely. But this will at least give you a sense for what needs to be done when you run a scan.

86
00:19:06.280 --> 00:19:20.789
Fran Zablocki: And then the last piece, if content is stale, expired, or critical, it may be time to update it, or redirect it, or retire it. And so, a lot of, conversations I have with content managers.

87
00:19:20.790 --> 00:19:32.159
Fran Zablocki: Are around just reducing the overall size of the website, and the number of pages and number of instances we have of things, because it makes it easier to manage.

88
00:19:32.160 --> 00:19:42.339
Fran Zablocki: And so running something like Content Intelligence can identify where you might be able to just cut out some older content that hasn't been performing well or hasn't gotten a lot of traffic.

89
00:19:46.350 --> 00:19:54.149
Joel Goodman: And we want to go on to another poll off of that. When was the last time you audited the content across your website?

90
00:19:54.230 --> 00:20:04.419
Joel Goodman: as we talk about Fran, like, the prioritization part is the most, difficult part within this, and so, you know, a lot of times these audits…

91
00:20:04.420 --> 00:20:17.220
Joel Goodman: take a long time, they take a lot of effort. And we see that a lot, and that's, again, why content intelligence exists. It helps us do these audits at scale across thousands and thousands and thousands of pages.

92
00:20:21.410 --> 00:20:26.200
Fran Zablocki: It'll be interesting to hear… How recently people have done this.

93
00:20:27.400 --> 00:20:35.909
Joel Goodman: Yeah, I'm curious, I mean, how often, or I guess how long have these projects traditionally taken for you when you've done, say, a large university content audit?

94
00:20:36.660 --> 00:20:42.530
Fran Zablocki: Well, I'm gonna make everybody feel a little bit better about themselves, hopefully, with this answer, because there are, in a lot of cases.

95
00:20:42.930 --> 00:21:02.609
Fran Zablocki: pages that have never been audited, because they weren't important enough, or there's just too many pages and not enough human effort behind them, and, you know, a lot of times, people only get to, like, the top 20 to 30% of their pages if it's a team that's strapped and has a lot of other responsibilities and a lot of things on their plates. So…

96
00:21:02.680 --> 00:21:10.860
Fran Zablocki: Yeah, I mean, some people get to it all, they have good systems set up, but I think, you know, the point we're trying to make across the board here is…

97
00:21:10.880 --> 00:21:23.899
Fran Zablocki: that sometimes it's just too much content for humans to be able to manage, and so that's what these results are kind of showing, too. We got kind of an even set of replies here. Good news is, majority of you,

98
00:21:23.950 --> 00:21:41.080
Fran Zablocki: or saying that you've had content audits in the last 3 months, that's great. But, you know, there's a good percentage of you, 20%, that are saying that they've never had a content audit, or none that they know of, and that's actually perfectly normal, too. I mean, we work a lot with folks who just…

99
00:21:41.230 --> 00:21:50.459
Fran Zablocki: haven't had a chance to do this, and that's exactly the reason why we've created Content Intelligence, because it can do, again, it can do that heavy lifting, and it can do a job that really nobody else has had time to do.

100
00:21:51.440 --> 00:22:11.020
Joel Goodman: And I think the important thing is that even if you have recently looked across all of your content and looked at whether it's performing or not, there's a big question of how long it's going to be up-to-date, you know? Will it go stale? How quickly will AI search not recognize it or not, not trust it as much?

101
00:22:11.020 --> 00:22:27.049
Joel Goodman: how many people have their hands and their keyboards built in there as well, right? So, we want to talk about good practice at scale, you know, meaning when you've got hundreds or thousands of pages, that's where it gets really difficult to maintain a grasp on all of this.

102
00:22:27.050 --> 00:22:44.879
Joel Goodman: You know, the loop's super clear on one page, or 10 pages, or maybe… maybe 100 pages if you've got the team to do it, but it does get, it does get difficult when you've got lots and lots and lots of pages or content spread across different teams and topics.

103
00:22:44.880 --> 00:23:00.420
Joel Goodman: You know, multiple pages answering the same question, different governance structures where, you know, owners don't know when to review content, or just aren't doing it because it's not set in. And then it gets really difficult because you don't know what to fix first.

104
00:23:00.570 --> 00:23:03.590
Joel Goodman: So it is a scale problem, and…

105
00:23:03.820 --> 00:23:18.900
Joel Goodman: What we would like to do, even, you know, a good number you've audited recently, we saw, but, you know, the real question again is, like, how long does that… that picture of your content stay accurate once that content starts to get moving again and more people start working on it?

106
00:23:20.200 --> 00:23:26.790
Joel Goodman: I want to show a video, and Fran and I will kind of talk through how this works, but…

107
00:23:26.790 --> 00:23:47.409
Joel Goodman: We built content intelligence around, looking at thousands and thousands and thousands of pages, and so we look across all of your content, and then we split it between accessibility and then also AI readiness, meaning, you know, can AI read your content? Are you doing all these things to give good trust signals out to the AI?

108
00:23:49.660 --> 00:23:56.209
Joel Goodman: So, accessibility is important because it, it's kind of the baseline, right, Fran? It's the baseline of whether.

109
00:23:56.210 --> 00:24:21.179
Joel Goodman: whether a machine can read your content in the first place. So we know it's important for, you know, for kind of equitable work, for making sure everyone can access your website, but the same tools that people with disabilities are using to read content on your site are also being used by AI technology to understand it. And so our tool both looks at it, it gives you a breakdown of the

110
00:24:21.180 --> 00:24:39.250
Joel Goodman: impact of the different infractions against WCAG 2.2, and then gives you a solution of how to fix these sorts of things. It wants… we want you to understand what the problems are and why they're important to fix, and then how to fix them.

111
00:24:40.020 --> 00:25:00.789
Fran Zablocki: Yeah, and the thing I really like that's showing right now is that it's finding the larger issue, pointing out the pages that that issue exists on, then really drilling down into, hey, this is the code in here that is not meeting that need, and here's a suggested piece of code to fix it. So that last bit where you saw the, kind of, the red and the green, was really, the part that I find most

112
00:25:00.790 --> 00:25:08.749
Fran Zablocki: interesting and effective for content managers. Now, on the AI readiness side, we like to say that on the accessibility side, this is, you know, making it accessible for humans.

113
00:25:08.750 --> 00:25:26.079
Fran Zablocki: And the AI readiness side is really making it accessible for AI in search. This is showing the Squiz website and showing the topics that our content intelligence tool has identified as, important, and is also showing our relative performance, even with

114
00:25:26.120 --> 00:25:39.329
Fran Zablocki: content that is being shown as AI-ready without high-priority issues. The system has a number of different other priorities, and so it's showing that we have 4 lower priority issues and about 30, 31 pages that are associated with that.

115
00:25:39.330 --> 00:25:50.839
Fran Zablocki: So you kind of get that high-level overview, but then you can drill down into a lot of detail, and I think one of the most valuable parts of the content intelligence is what it's showing now, which is just how many questions it's asking of each topic.

116
00:25:50.860 --> 00:26:10.189
Fran Zablocki: it's emulating the same kind of behavior that AI Search does, which is called fan-out querying, and it's really hitting your website with as many possible variations on questions as it can think of, and you have full insight and visibility into those questions and can kind of see how it's answering them on your behalf.

117
00:26:10.190 --> 00:26:16.370
Fran Zablocki: This is a good example of, like, drilling all the way down into an individual issue, and it's pointing out

118
00:26:16.600 --> 00:26:27.560
Fran Zablocki: exactly what's needed for that particular page, and making a suggestion. This is showing how we're using our own tool on our website,

119
00:26:27.930 --> 00:26:47.390
Fran Zablocki: conversational search is something that has become very, very common for all of us to use through things like ChatGPT, but we can now enable this exact same kind of experience on our website or your website. It comes up with answers in written text, but it also does citations of all the different sources.

120
00:26:47.390 --> 00:26:51.650
Fran Zablocki: And is able to ask follow-up questions so you can chain together

121
00:26:51.650 --> 00:26:58.839
Fran Zablocki: Different, follow-up details, and kind of retain that conversation, throughout.

122
00:27:00.090 --> 00:27:11.929
Joel Goodman: What's great is that all those questions that Content Intelligence uses to ask and interrogate your content are used to make the results better, and more accurate within our conversational search product.

123
00:27:11.930 --> 00:27:27.040
Joel Goodman: And, that's one of those things that means we basically eliminate hallucinations entirely, because it's only answers coming from your content. We're not making anything up. In fact, the model doesn't even have access to make something up in that sort of a way.

124
00:27:29.870 --> 00:27:33.410
Joel Goodman: Yes, with that… Oh, go ahead, Fran.

125
00:27:34.020 --> 00:27:50.820
Fran Zablocki: Yeah, I was just gonna say, I really like conversational search and content intelligence paired together, because content intelligence is kind of the blanket, audit tool to make all of your content better, and it's… it is the source, it's improving the source so that all the different destinations

126
00:27:50.820 --> 00:28:11.800
Fran Zablocki: Are also improved, right? The destination we just showed, which would be, like, your local search that's, you know, in your own conversational search, or even keyword search, is going to be improved. The different, AI models are going to have more visibility. You're going to have more visibility with those different AI models. It's going to improve accessibility scores, which is going to improve for people who have,

127
00:28:12.030 --> 00:28:24.789
Fran Zablocki: screen readers, and it's also, going to improve regular old SEO, because it's really looking at, you know, what kind of keywords are being used. So, it kind of covers all the different modes in a really easy-to-use interface.

128
00:28:26.440 --> 00:28:45.419
Joel Goodman: Content Intelligence works on any website, too. You don't have to be on the full SquizDXP, and it's really meant to help you get a handle on your content by helping you improve AI search visibility, auditing your website for accessibility, helping you prioritize what needs to be fixed, and even telling you what

129
00:28:45.420 --> 00:28:49.440
Joel Goodman: What needs to be fixed and how to fix it, and also where to fix it.

130
00:28:49.440 --> 00:29:04.669
Joel Goodman: So, we are… we're offering, a one-month trial. If you would, like to check it out, you can… you can hit that QR code, or we will, well, hit the… hit the link in the chat, and this will also come out.

131
00:29:04.720 --> 00:29:20.500
Joel Goodman: with the email that we sent afterwards. We'd love to show you how this works in person. You saw the quick demo across what we've done with Squiz.net, but we can actually show you that live, and you can ask questions, so we can dig into some of the things that didn't show up in that demo video.

132
00:29:22.830 --> 00:29:43.090
Joel Goodman: Key takeaways today, again, just to recap what we talked about, AI search is magnifying existing content health problems. You see that all the time with other things you use AI for. It always kind of amplifies the good and the not-so-good. AI depends on strong structure, metadata, authority, and freshness to understand your content and know what to trust.

133
00:29:43.230 --> 00:30:01.510
Joel Goodman: And a repeatable loop can keep those signals to AI search strong. And Squiz Content Intelligence can help you make that loop repeatable at scale, so you aren't having to cobble it together yourself out of, out of what you have, at hand, which is… tends to be spreadsheets and…

134
00:30:02.190 --> 00:30:13.510
Joel Goodman: notepads and things like that. So, with that, we are happy to take some questions. If you want to drop some questions in the Q&A, looks like we've got a couple that have popped up here.

135
00:30:13.510 --> 00:30:25.540
Joel Goodman: And fran, I want to go back to… I wanna go back to, what do you… what do you think the best way to decide which pages need attention first? Like, you talked about the analytics side of it.

136
00:30:25.540 --> 00:30:27.989
Joel Goodman: And, and kind of the traffic.

137
00:30:28.010 --> 00:30:39.789
Joel Goodman: portions, like, kind of looking at it that way, I mean, how do you think an organization should decide, kind of, that high-priority, leverage piece for which content to tackle first?

138
00:30:40.420 --> 00:30:44.920
Fran Zablocki: Yeah, yeah, so I think there's two halves to it. I do think that analytics are important, but…

139
00:30:45.110 --> 00:31:00.750
Fran Zablocki: analytics will not answer the question of why pages that are really important aren't getting the traffic that they should, right? So you… you may have pages that you really want there to be a lot of traffic to, or expect there to be a lot of traffic to, and it's not showing up in analytics. And so.

140
00:31:00.750 --> 00:31:05.709
Fran Zablocki: What you need to do is cross-reference what you're seeing in analytics with what your key…

141
00:31:05.710 --> 00:31:26.929
Fran Zablocki: customer or student or visitor journeys are, and mapping out those journeys across, the website. So, for example, let's take higher education. Somebody who is trying to decide whether or not they want to come to your school is going to need to look at a number of key areas of the website. And so, by look… by identifying what the key

142
00:31:26.940 --> 00:31:35.610
Fran Zablocki: touchpoints are throughout that journey, and the pages that are really answering those key questions that they might have, and also providing those key conversions, whether it's applying or…

143
00:31:35.710 --> 00:31:38.900
Fran Zablocki: You know, contacting somebody on campus.

144
00:31:38.950 --> 00:31:46.800
Fran Zablocki: you can plot out the pages that are most important to that conversion and to that audience. And so, that's a list.

145
00:31:46.810 --> 00:32:02.990
Fran Zablocki: that you have, then, that's a priority list, and you can cross-reference it with your analytics, and that can actually identify when you have pages that are really, really key and important for people to find, but people are not finding it. And then using something like Content Intelligence.

146
00:32:02.990 --> 00:32:12.929
Fran Zablocki: You'll be able to look at topically, okay, are we actually answering the questions associated with this piece, and cross-reference it as a third signal to see whether or not

147
00:32:12.930 --> 00:32:24.070
Fran Zablocki: those questions are being answered at all, and if they are being answered, kind of what's being sourced there? Like, are they the pages that you expect? So, I think using all three of those inputs can really help

148
00:32:24.070 --> 00:32:36.159
Fran Zablocki: in terms of prioritization, in addition to what content intelligence is going to do, which is to, you know, prioritize things as high, medium, and low on its own. So it's going to give some suggestions there. So you've got a lot of different signals and inputs there.

149
00:32:36.160 --> 00:32:40.789
Fran Zablocki: To figure that out, but start with your customer journey, start with the core things that people

150
00:32:40.790 --> 00:32:46.929
Fran Zablocki: That you know people must do on your site for it to be considered successful, and then use those other tools and signals.

151
00:32:48.640 --> 00:33:06.340
Joel Goodman: Ben asks, do you need SquizDXP for full functionality of Content Intelligence? And, Content Intelligence works on whatever stack you have, so you can… you don't have to be on SquizDXP managing your content there. Down the line, there will be some things that…

152
00:33:06.340 --> 00:33:16.299
Joel Goodman: you know, that probably the product will work better with it, you know, we'll need, you know, things like updating content automatically or things like that, you may need the DXP for.

153
00:33:16.300 --> 00:33:22.530
Joel Goodman: But right now, it works on your site, and it's meant to work on any site that's there.

154
00:33:24.850 --> 00:33:45.980
Joel Goodman: As mentioned, second half to that question, earlier it was mentioned that the search feature wasn't capable of hallucinating as a certainty. It's working off of the content that's on your site. If it can't find an answer within the content that's on your site, it is going… it's designed to say, I don't have a good answer for that. You can actually decide what that exact message is that pops up.

155
00:33:45.980 --> 00:33:54.649
Joel Goodman: But, the only way that it can give a poor answer is if the content hasn't been updated, and that's just how large language models work.

156
00:33:54.650 --> 00:34:10.449
Joel Goodman: And so that's why you have to use the content intelligence tool to improve your content, update, and make sure that it's not confusing to a model, but it's only going off of the content that's on your site. If it can't find an answer in the content on your site, it's not going to make something up.

157
00:34:12.480 --> 00:34:28.959
Joel Goodman: And then, Hannah asks if you just use keyword and boolean search terms in conversational search. Not really, it's built to split out a keyword into separate keyword search, but conversational search itself is fully AI.

158
00:34:28.960 --> 00:34:42.059
Joel Goodman: But we can, we can talk through details, with you on how that works. We still offer our keyword search with Funnelback, and conversational search is, is another layer on top of that.

159
00:34:45.170 --> 00:34:52.540
Joel Goodman: How can we write clearly for AI without losing our organization's tone of voice or sounding too generic?

160
00:34:53.199 --> 00:34:54.739
Fran Zablocki: This is an interesting one.

161
00:34:55.000 --> 00:34:59.600
Fran Zablocki: Yeah, this is an interesting one. You don't want to…

162
00:34:59.780 --> 00:35:07.509
Fran Zablocki: sanitize for AI so much that you lose that, sense of personality and that tone, and voice.

163
00:35:07.650 --> 00:35:12.179
Fran Zablocki: I think it's a balance. I think that…

164
00:35:12.660 --> 00:35:32.190
Fran Zablocki: we… well, we know that AI really skims over and kind of ignores marketing fluff in general, right? Like, it's looking for material answers to the questions that are asked, it's looking for referential information, it's looking for procedural information, it's looking for specific details,

165
00:35:32.350 --> 00:35:33.350
Fran Zablocki: Because…

166
00:35:33.690 --> 00:35:44.970
Fran Zablocki: people are asking very detailed, multi-part questions that are a site of marketing copy. It just means that you have to balance marketing copy bylines, and things like that.

167
00:35:45.060 --> 00:35:56.469
Fran Zablocki: with having real substantive information on those pages as well. So, I think there's still room, and there's a need for experiential

168
00:35:56.700 --> 00:36:10.049
Fran Zablocki: content on the website for people to understand who you are, and get a feel for who you are, and what you can do for them, and get a feel for, you know, your ethos of the institution. And there is a space for marketing language, but I do think that

169
00:36:10.830 --> 00:36:29.609
Fran Zablocki: sites that are overly reliant on just marketing language, and are not providing substantive answers are gonna start suffering in terms of their results, and so I think it's gonna be a matter of striking a balance. And you can still also include your tone and style within those substantive informational answers.

170
00:36:29.610 --> 00:36:34.440
Fran Zablocki: They just won't, you know, necessarily have the level of marketing

171
00:36:34.510 --> 00:36:41.960
Fran Zablocki: Language there, but, you know, you want to still have consistency across all of your content, so it sounds like it's being written from one source.

172
00:36:44.150 --> 00:36:54.110
Joel Goodman: Fran Ben asks, how does AI interact with PDFs within the website, and is it best practice to put any information from PDFs into a page, into page text?

173
00:36:55.830 --> 00:36:57.820
Fran Zablocki: It is best practice to put

174
00:36:57.820 --> 00:37:22.810
Fran Zablocki: information from PDFs into page text. I mean, this has been a challenge for a long time. PDFs, by default, unless you put a decent amount of work into them, and even then, they're not ideal, are not accessible, so they're not getting picked up by screen readers in a lot of cases. They're not great for SEO, and they're not great for humans, either. People are looking to be able to read something within HTML and responsive design and not have to, you know, pinch and zoom

175
00:37:22.810 --> 00:37:41.099
Fran Zablocki: Zoom on a mobile phone for that experience. So, it is best practice to put them into PDFs for AI as well. AI is not going to be able to pull out that information. AI needs it to be within HTML. You know, it may be able to surmise some detail about the PDF.

176
00:37:41.100 --> 00:37:44.270
Fran Zablocki: But it's no guarantee. So, and, and just like…

177
00:37:44.340 --> 00:37:53.990
Fran Zablocki: in every other mode, we really need to make sure that we've got an HTML version that can be crawled and picked up. If you need to have a PDF,

178
00:37:53.990 --> 00:38:08.800
Fran Zablocki: Because it's something that makes sense, like, someone's going to need to have a printable piece for this that they can carry around. You can have a download for that, but it shouldn't be an either-or situation. Anytime you have a PDF full of content, there should be an equivalent HTML version of it.

179
00:38:09.740 --> 00:38:23.660
Joel Goodman: Yeah, PDFs are risky, I think. They're just risky from another… from a bunch of different standpoints. They're risky from an accessibility standpoint. In a lot of countries, accessibility compliance PDFs kind of hit up against that because they're largely inaccessible.

180
00:38:23.660 --> 00:38:34.840
Joel Goodman: And for any of you that have been going through a massive PDF-to-web migration project, you understand the pain of

181
00:38:34.840 --> 00:38:49.089
Joel Goodman: the choices of choosing PDF 10, 15 years ago for the format for your content and moving back the other way. So I think they're extremely risky, and we know they don't perform as well because they don't show up in as many references, and…

182
00:38:49.110 --> 00:38:51.570
Joel Goodman: They are difficult for screen readers to read.

183
00:38:54.090 --> 00:38:58.559
Joel Goodman: Alright, one more here,

184
00:38:58.840 --> 00:39:05.509
Joel Goodman: Fran, what are the clearest signs that a piece of content should be consolidated, redirected, or retired?

185
00:39:08.350 --> 00:39:11.499
Fran Zablocki: Well, like I mentioned earlier, I think you can…

186
00:39:11.990 --> 00:39:28.199
Fran Zablocki: just take a look at the published date of something and start to get to some of the low-hanging fruit here, right? So, I mean, looking at your analytics and seeing what your oldest content is, and starting from that side, if it's something that is over a year old.

187
00:39:28.320 --> 00:39:34.420
Fran Zablocki: And it also does not get any significant traffic. It is a prime candidate to be removed.

188
00:39:34.450 --> 00:39:37.230
Fran Zablocki: From the site. Retired from the site.

189
00:39:37.230 --> 00:39:54.510
Fran Zablocki: If it is an older piece of content that has a lot of traffic, then you can make an exception to that. Like, maybe it's an article that somebody wrote years ago that still has a lot of traction, that still gets a lot of traffic, so it's not a cut and dry, just get rid of everything that's over a year old, but combining the age plus

190
00:39:54.510 --> 00:40:13.150
Fran Zablocki: how much viewership it gets can give you an idea for what can easily be removed. And then, you know, from there, I think it's also worth just taking a look at how much content is on a particular page. Like, one of the things that I always do is just see if a page even needs to be a page. You know, if there's just not enough

191
00:40:13.240 --> 00:40:15.220
Fran Zablocki: Content on a page,

192
00:40:15.260 --> 00:40:34.229
Fran Zablocki: to warrant it being its own page, then it's prime for consolidation into a larger page, and vice versa. You might have pages that just are trying to do too much, and trying to, like, lift too many topics all at once, and in that case, we're gonna go in the opposite direction, and recommend that you chop that up into dedicated pages.

193
00:40:34.230 --> 00:40:46.910
Fran Zablocki: And that gets back to that one signal of, you know, being, like, the authoritative source on things. If you're trying to do too much with a single page, it can confuse everybody, humans included, as to what, you know, where the primary destination for that is.

194
00:40:48.950 --> 00:40:59.979
Joel Goodman: And with that, we'd like to say thank you so much for coming to this webinar. Again, if you would like to trial Squiz Content Intelligence, you can hit that

195
00:40:59.980 --> 00:41:13.710
Joel Goodman: QR code of the link in the chat, but we can also offer you a free AI visibility report, for your specific website over a limited, set of content that you have now. You can hit that new link that just dropped in the chat.

196
00:41:13.710 --> 00:41:30.140
Joel Goodman: If you've got any other questions, feel free to follow up by email, the email that you got this, this registration form from. Please, please send us any questions, we're happy to chat with you. And keep an eye out for the next webinars and also the recording in your inbox soon.

197
00:41:30.160 --> 00:41:32.740
Joel Goodman: Thank you so much for joining us today.

198
00:41:33.210 --> 00:41:35.520
Fran Zablocki: Pleasure spending some time with you. Thank you.