Latest / Found in AI: AI Search Visibility, SEO, & GEO / AI Search News: Google's Latest AI Mode Test & New ChatGPT Visibility Research

Transcript

AI Search News: Google's Latest AI Mode Test & New ChatGPT Visibility Research · Found in AI: AI Search Visibility, SEO, & GEO

• 0:41
  SPEAKER_00: Hey and welcome back to Found in AI, the podcast where we talk about getting your brand found in AI Search.
• 0:47
  SPEAKER_00: I'm Cassie Clark, an AI Search Visibility Consultant.
• 0:51
  SPEAKER_00: Today is October 29th and I have three updates to cover.
• 0:55
  SPEAKER_00: Google assessing a new way to display website links in AI mode.
• 1:00
  SPEAKER_00: A study from Graphite suggests that even when ChatGPT cites pages that mention your brand, it might still leave you out of the answer.
• 1:08
  SPEAKER_00: And Search Bloom has some new data about just how much AI visibility happens on websites brands don't actually own.
• 1:16
  SPEAKER_00: Let's get into it.
• 1:25
  SPEAKER_00: First up, Google is experimenting with how it displays organic website links in AI mode.
• 1:32
  SPEAKER_00: Search Engine Watch reported earlier this week that Google is testing a section called Further Exploration Under AI Generated Answers.
• 1:41
  SPEAKER_00: Instead of relying solely on hyperlinks within the answer itself, this section displays website links with descriptions, images, and dates.
• 1:50
  SPEAKER_00: Now, this is really just a test.
• 1:52
  SPEAKER_00: We do not know how widely it's being shown, whether Google intends to roll it out permanently, or even exactly what those dates represent.
• 2:00
  SPEAKER_00: It could be publication dates, update dates, or something else, but this is interesting.
• 2:06
  SPEAKER_00: One of the ongoing questions about AI search is what happens to websites providing the information.
• 2:11
  SPEAKER_00: We have talked about this plenty on the podcast.
• 2:15
  SPEAKER_00: If Google answers the user's question directly, what incentive does the user have to actually just click through?
• 2:23
  SPEAKER_00: And that's why the presentation of those sources matters.
• 2:26
  SPEAKER_00: A link buried in an AI-generated paragraph is a very different experience from a dedicated section from a website with a name, image, and description.
• 2:37
  SPEAKER_00: Does that mean this particular design will increase clicks?
• 2:41
  SPEAKER_00: Well, we don't know that.
• 2:42
  SPEAKER_00: Nobody should be making that claim based on a screenshot of just one test.
• 2:47
  SPEAKER_00: But it does suggest that Google is still experimenting with how to connect AI-generated answers to the wider web.
• 2:54
  SPEAKER_00: And from a marketing perspective, I think that's worth paying attention to.
• 2:58
  SPEAKER_00: If Google starts giving websites more prominent placements alongside AI-generated answers, brands will need to understand what information gets displayed and then how those placements affect actual referral traffic.
• 3:11
  SPEAKER_00: For now, I don't think I would change my content strategy or marketing strategy over this.
• 3:16
  SPEAKER_00: I would just keep an eye on it though, because how AI Search displays a source could eventually matter almost as much as whether it cites that source at all.
• 3:25
  SPEAKER_00: Which brings me to the second story.
• 3:28
  SPEAKER_00: Search Engine Journal published an interesting breakdown this week of research from Graphite.
• 3:33
  SPEAKER_00: Graphite analyzed how often ChatGPT mentioned brands in comparison style questions, and they looked at responses with and without web search.
• 3:42
  SPEAKER_00: The researchers found a relationship between how often a brand appeared in answers without search and how often it appeared when ChatGPT ultimately did use search.
• 3:53
  SPEAKER_00: Among brands mentioned on at least 15% of the pages Chat GPT cited, those appearing in fewer than a quarter of the no search answers had a medium visibility of just 11% when search was used.
• 4:09
  SPEAKER_00: So, in other words, ChatGPT could retrieve and cite your pages discussing a brand and still not include that brand in its answer.
• 4:19
  SPEAKER_00: So if you've been listening to the podcast or you've been following me on LinkedIn the last little bit, what I'm about to say next sounds familiar.
• 4:31
  SPEAKER_00: And these engines pull information from a ton of different places when constructing an answer.
• 4:37
  SPEAKER_00: Graphite's interpretation of all of this is that brands should pay attention not only to the information ChatGPT retrieves from the broader web, but also to what the model already appears to know.
• 4:49
  SPEAKER_00: The agency calls this parametric knowledge optimization, or PKO.
• 4:55
  SPEAKER_00: Essentially, they are arguing that information learned during model training may influence which brands ChatGPT considers when it generates an answer.
• 5:06
  SPEAKER_00: I think that is a reasonable hypothesis to investigate from this research, but I also really want to be careful about how we interpret the findings.
• 5:16
  SPEAKER_00: First, this is observational research.
• 5:19
  SPEAKER_00: It shows a relationship between brand mentions with and without search, but it does not establish that changing what a model learned during training will cause it to recommend a brand more often.
• 5:30
  SPEAKER_00: Second, the study doesn't fully account for existing brand awareness.
• 5:34
  SPEAKER_00: A well-known company may be more likely to appear both in training material and on the websites ChatGPT retrieves.
• 5:41
  SPEAKER_00: That alone could help some of the relationship.
• 5:44
  SPEAKER_00: And third, marketers have very limited control over what ultimately gets included in a model's training data.
• 5:52
  SPEAKER_00: We can publish useful content, we can make sure our websites are accessible to the appropriate crawlers, and we can build a recognizable brand that gets discussed across the web.
• 6:04
  SPEAKER_00: But we simply can't submit a page and then guarantee that a future model will learn it, remember it, or even use it in a recommendation.
• 6:12
  SPEAKER_00: So there are two things to take away from this research.
• 6:15
  SPEAKER_00: First, if you are only measuring citations, you are probably missing a big part of the picture.
• 6:22
  SPEAKER_00: You also need to measure whether your brain is being named in the answer, how it's being described, and whether it's being recommended for the questions that matter to your customers.
• 6:32
  SPEAKER_00: Think about those money prompts.
• 6:34
  SPEAKER_00: And because those answers vary, you need to repeat these measurements.
• 6:39
  SPEAKER_00: And second, everything is training data.
• 6:43
  SPEAKER_00: Every blog post, social media post, video, comment, everything.
• 6:49
  SPEAKER_00: But like I just said, you have very limited control over what actually gets included in a model's training data.
• 6:56
  SPEAKER_00: So just keep that in mind as you work on your marketing strategy.
• 7:01
  SPEAKER_00: Okay, our third update.
• 7:03
  SPEAKER_00: Search Bloom released a study yesterday looking at where citations appeared across its AI visibility tracking.
• 7:10
  SPEAKER_00: The agency analyzed over 68,000 AI answers covering 146 questions across eight engines over 90 days.
• 7:21
  SPEAKER_00: The big number from this research is 91.1% of the cited sources were third-party pages.
• 7:30
  SPEAKER_00: Only 8.9% came from SearchBloom's own website.
• 7:34
  SPEAKER_00: The most cited source, it was a directory.
• 7:38
  SPEAKER_00: Now, before everybody starts reallocating 90% of their content budgets to third-party websites, we need to talk about the methodology here.
• 7:48
  SPEAKER_00: SearchBloom used its own website as the test case, and many of the questions were about finding or hiring an agency, including searches along the lines of who's the best or who should I hire.
• 8:02
  SPEAKER_00: Think about the types of websites that tend to answer those questions.
• 8:05
  SPEAKER_00: It could be agency directories, review platforms, comparison list, or industry roundups.
• 8:12
  SPEAKER_00: So it isn't especially surprising that third-party websites made up a large share of citations for that particular set of questions.
• 8:21
  SPEAKER_00: And it doesn't mean that 91% of all AI citations across every industry will come from third-party websites.
• 8:30
  SPEAKER_00: So I wouldn't take that number as gospel and then use it as a universal benchmark, but I do think that the broader point here is worth discussing.
• 8:39
  SPEAKER_00: A lot of companies approach AI visibility by looking almost exclusively at their own website.
• 8:45
  SPEAKER_00: They want to know whether their pages are structured correctly, whether they've implemented a schema markup, and whether their content is optimized for the questions people ask those AI engines.
• 8:57
  SPEAKER_00: And yes, don't think that is me saying those things don't matter.
• 9:00
  SPEAKER_00: They absolutely do.
• 9:04
  SPEAKER_00: But your website is only just one part of the information environment surrounding your brand.
• 9:10
  SPEAKER_00: You can publish the most beautifully structured product page in the world.
• 9:15
  SPEAKER_00: But if someone asks an AI engine which provider they should hire, that tool tends to also look at those independent sources for comparisons and recommendations.
• 9:27
  SPEAKER_00: And this is where I think the SearchBloom study puts things into perspective.
• 9:32
  SPEAKER_00: Not because 91.1% is some sort of magic number every brand should chase, but because it reinforces the importance of understanding where your brand visibility is actually coming from.
• 9:46
  SPEAKER_00: If you find that AI answers in your category consistently cite certain publications, directories, or review platforms, it gives you some really helpful information about your broader marketing strategy and maybe where you should put your focus.
• 10:01
  SPEAKER_00: You need brand building, public relations, reputation management, and content strategy working together.
• 10:09
  SPEAKER_00: Which, if you have listened to this podcast for any length of time, that is something that I keep coming back to.
• 10:16
  SPEAKER_00: So let's put these three updates together.
• 10:19
  SPEAKER_00: Google is experimenting with how websites appear alongside AI-generated answers.
• 10:25
  SPEAKER_00: Graphite's research suggests that being mentioned on a cited page doesn't guarantee your brand will make it into the answer.
• 10:33
  SPEAKER_00: And SearchBloom study highlights how important third-party websites can be in the sources that AI engines use.
• 10:41
  SPEAKER_00: To me, the takeaway here is just it's pretty straightforward.
• 10:44
  SPEAKER_00: Getting cited, getting mentioned, and getting recommended are three different things.
• 10:50
  SPEAKER_00: And if your AI search strategy only focuses on just one part of that, you're not seeing the complete picture.
• 10:58
  SPEAKER_00: Okay, that's all I've got for you on today's update.
• 11:01
  SPEAKER_00: If you found this episode helpful, make sure you are following Found in AI.
• 11:05
  SPEAKER_00: And if you want more analysis of what's happening in AI Search, you can subscribe to my newsletter, The Visibility Report.
• 11:12
  SPEAKER_00: It goes out every Friday, and the link is in the show notes.
• 11:15
  SPEAKER_00: And one last thing before I let you go.
• 11:18
  SPEAKER_00: I'm posting a free live training next Thursday, November 5th at 1 p.m.
• 11:23
  SPEAKER_00: Eastern Time.
• 11:24
  SPEAKER_00: It's called How to Build an AI Search Visibility Strategy.
• 11:28
  SPEAKER_00: We'll talk about how to approach AI Search strategically, where to focus your efforts, and how to measure whether those efforts are actually working.
• 11:36
  SPEAKER_00: I will also have time for a few live questions.
• 11:39
  SPEAKER_00: You can save your seat over at Cassie Clarkmarketing.com slash AI Search Training.
• 11:44
  SPEAKER_00: And as always, that link is in the show notes as well.
• 11:48
  SPEAKER_00: Okay, that's it for today.
• 11:49
  SPEAKER_00: Until next time, stay visible.