Latest / Found in AI: AI Search Visibility, SEO, & GEO / Microsoft Clarity's AI Visibility Metrics and Google's Reddit Denial

Transcript

Microsoft Clarity's AI Visibility Metrics and Google's Reddit Denial · Found in AI: AI Search Visibility, SEO, & GEO

• 0:41
  SPEAKER_00: Welcome back to Found in AI.
• 0:43
  SPEAKER_00: I am Cassie Clark, an AI Search Visibility Consultant, and this is the Thursday news update where we go through what actually happened this week in AI Search.
• 0:51
  SPEAKER_00: Today is Thursday, August 6th, and I'm gonna be honest with you, it was a quiet week on the AI Search front.
• 0:57
  SPEAKER_00: There are only two stories that we're covering today, both of them are pointing at the same thing but from opposite directions.
• 1:03
  SPEAKER_00: Let's dig right in.
• 1:13
  SPEAKER_00: Okay, our first story is about Microsoft, and I've said this before, and I'll say it again, they are truly the pioneers when it comes to AI search visibility.
• 1:21
  SPEAKER_00: So if anyone at the Microsoft team is listening, big fan.
• 1:25
  SPEAKER_00: Microsoft Advertising launched a product newsletter on LinkedIn this week, and in the first edition, there is one line that I have read a couple of times now.
• 1:33
  SPEAKER_00: Here's what it says We heard you that the AI era demands the same rigor and transparency of reporting.
• 1:39
  SPEAKER_00: This is why we've invested heavily to bring new AI visibility insights through Microsoft Clarity.
• 1:45
  SPEAKER_00: The AI era demands the same rigor and transparency of reporting.
• 1:50
  SPEAKER_00: Okay, well, same as what?
• 1:52
  SPEAKER_00: Same as paid search.
• 1:53
  SPEAKER_00: The same as the reporting standard advertisers have had for 20 years, and organic has never really quite matched.
• 2:01
  SPEAKER_00: That's a platform kind of saying out loud that AI visibility is now a reportable service with a reporting standard.
• 2:08
  SPEAKER_00: It's the kind of thing you're expected to be able to show your CFO.
• 2:12
  SPEAKER_00: So, what's actually in it?
• 2:14
  SPEAKER_00: Well, Microsoft's Clarity's AI Visibility Suite now includes something called Topic Insights, which tells you how your content is understood by AI, which topics AI associates with your brand, and your inclusion rate.
• 2:26
  SPEAKER_00: And alongside that, they call out four specific metrics.
• 2:29
  SPEAKER_00: So let's cover those.
• 2:30
  SPEAKER_00: First is citations, or when a page of yours is referenced as a source in an AI-generated answer after being retrieved during grounding.
• 2:38
  SPEAKER_00: They also call out citation share, which is how often your domain is cited for a given grounding query.
• 2:44
  SPEAKER_00: They call out grounding queries or the retrieval queries that the AI system generates behind the scenes to find what it needs to answer a user's prompt.
• 2:53
  SPEAKER_00: So not what the user typed, but what the machine went looking for.
• 2:57
  SPEAKER_00: And then they also mention share of authority or the percentage of total citations attributed to your domain compared with every other cited domain in the same answer set.
• 3:07
  SPEAKER_00: Share of authority is a platform-owned, platform-defined free metric for how AI systems pick you over your competitors.
• 3:15
  SPEAKER_00: There is an entire category of AI visibility tools that exist to measure roughly that, and they charge for it.
• 3:23
  SPEAKER_00: Microsoft just shipped a version of this inside a free analytics product and wrote the definition themselves.
• 3:30
  SPEAKER_00: When a platform defines a metric, the platform decides what counts.
• 3:34
  SPEAKER_00: That is not a criticism of anything else, it's really just the situation.
• 3:38
  SPEAKER_00: Anyone who spent time with Search Console really knows what I'm talking about here.
• 3:43
  SPEAKER_00: The data is useful and it's also exactly the data that Google chose to show you.
• 3:48
  SPEAKER_00: But here's the part that I actually want to talk about.
• 3:51
  SPEAKER_00: This went out in a paid media newsletter.
• 3:54
  SPEAKER_00: So if you read the piece, and I'll link it in the show notes, read all of that with that in mind.
• 4:01
  SPEAKER_00: Microsoft is now telling you which topics AI associates with your brand, what your inclusion rate is, what queries the model generated to go find you.
• 4:09
  SPEAKER_00: And then the recommended actions are check your search terms against grounding queries, adjust keywords and negative keywords, adapt ad creative and landing pages based on competitive intelligence.
• 4:21
  SPEAKER_00: That, my friends, is entity data handed straight to the PPC team to buy better ads with.
• 4:28
  SPEAKER_00: I'm not saying that's wrong.
• 4:29
  SPEAKER_00: Those are useful things for the PPC team to have.
• 4:32
  SPEAKER_00: But if your company installs Clarity this quarter, that dashboard is going to go live with whoever owns the Microsoft ad account.
• 4:40
  SPEAKER_00: And how AI systems describe your brand is not really a paid media question.
• 4:46
  SPEAKER_00: That's a positioning question, a PR question, an editorial question, and a product marketing question.
• 4:53
  SPEAKER_00: And the answer is going to sit in a tool that none of those people log into.
• 4:57
  SPEAKER_00: I'm going to keep repeating this.
• 5:00
  SPEAKER_00: This is the organization alignment problem with a product launch attached to it.
• 5:04
  SPEAKER_00: The data doesn't fix the fact that nobody owns the outcome.
• 5:09
  SPEAKER_00: If you take one action away from this episode, it should be this.
• 5:13
  SPEAKER_00: Find out whether anyone at your company has clarity installed, and then find out who's looking at it.
• 5:19
  SPEAKER_00: Try to argue to get that data shared with everyone.
• 5:21
  SPEAKER_00: Now, if you go read the newsletter, I do want to call out a tiny side note on the numbers.
• 5:26
  SPEAKER_00: The newsletter has some big stats in it, like a 53% higher conversion rate for ads surveying in AI experiences, 25% more relevant, an 8% lift in incremental conversions from PMAX, and a 45% higher conversion rate from search partners.
• 5:44
  SPEAKER_00: Now those are all Microsoft's numbers in Microsoft's ad product newsletter about Microsoft's ad products, and there's really no methodology attached to it.
• 5:54
  SPEAKER_00: I'm not calling any of that wrong.
• 5:55
  SPEAKER_00: I'm just saying that they're marketing claims until somebody shows the work.
• 5:59
  SPEAKER_00: And if those numbers mean anything to you, you should carry them as that, marketing claims, into any meeting where you're using them to justify your spend.
• 6:07
  SPEAKER_00: There's also one limitation that's worth calling out too.
• 6:10
  SPEAKER_00: What clarity shows you is how being in co-pilot's grinding process retrieves and understands your content.
• 6:16
  SPEAKER_00: That's a real signal about a real layer.
• 6:19
  SPEAKER_00: It's not the same as what a specific logged in user with months of accumulated chat history actually gets recommended when they ask about your category.
• 6:27
  SPEAKER_00: Those are different surfaces and they don't always agree.
• 6:30
  SPEAKER_00: Clarity brings some light to that retrieval layer, but the personalization layer is still kind of dark, and I'm gonna have more to say about that pretty soon, so stay tuned.
• 6:38
  SPEAKER_00: One more teeny tiny thing about this, and I want to flag it as an observation, not necessarily claim, because I have no idea what Microsoft's intent was, but I found it interesting.
• 6:47
  SPEAKER_00: The newsletter went out as a LinkedIn newsletter through LinkedIn Pulse.
• 6:52
  SPEAKER_00: Not a blog post, but a LinkedIn newsletter with a comment section asking readers to leave their questions in the comments for a future edition.
• 6:59
  SPEAKER_00: Now, LinkedIn is a Microsoft property, yes.
• 7:02
  SPEAKER_00: So the simplest explanation is really the obvious one.
• 7:06
  SPEAKER_00: They use the platform they owned the way any of us would.
• 7:09
  SPEAKER_00: But, and I keep overthinking this, like that's just how my brain works.
• 7:13
  SPEAKER_00: We also know that LinkedIn gets cited in AI answers pretty heavily.
• 7:17
  SPEAKER_00: It is a high trust domain with a clean structure, real author attribution, and consistent publishing.
• 7:24
  SPEAKER_00: That's every reason a model has to trust a source.
• 7:27
  SPEAKER_00: LinkedIn has that.
• 7:28
  SPEAKER_00: So, a company that just launched a product for measuring which sources AI systems select, publishing its product announcements on a platform that AI systems tend to select with an engagement mechanic that generates additional on-page content in the comments.
• 7:46
  SPEAKER_00: Do you see where I'm going with this one?
• 7:48
  SPEAKER_00: Maybe it's a coincidence, it might be, but it's the kind of coincidence I'd want to think about for my own distribution.
• 7:55
  SPEAKER_00: And again, I have no idea the reasoning for using LinkedIn Pulse for this, but it leads to the point that I care about, which holds up whether or not Microsoft fit any of this.
• 8:05
  SPEAKER_00: Newsletter content is worth something now, in a way that it wasn't really worth maybe even three years ago.
• 8:11
  SPEAKER_00: For a long time, the argument for a newsletter was audience ownership.
• 8:15
  SPEAKER_00: You own the list, the algorithm can't take that from you.
• 8:18
  SPEAKER_00: If the platform shuts down, you still have your list.
• 8:20
  SPEAKER_00: That is all still true.
• 8:22
  SPEAKER_00: It is still a good reason to have your own list, but there's a second reason now, and it's about where the content actually lives.
• 8:29
  SPEAKER_00: A newsletter published on a platform with a public archive is indexable, citable, structured, consistently dated, and attributed to a named human being.
• 8:40
  SPEAKER_00: It hits freshness, it hits structure, and if you're publishing under a person's name, it feeds entity authority.
• 8:46
  SPEAKER_00: That's the whole FSA framework in one asset.
• 8:49
  SPEAKER_00: A newsletter that only ever arrives in an inbox is a relationship.
• 8:53
  SPEAKER_00: But a newsletter that also lives on a public, crawlable, trusted archive is a relationship and a retrieval source.
• 9:01
  SPEAKER_00: If you're running a newsletter that exists exclusively in email, which might be good for the Google surfaces and the personal intelligence thing they have going on, I would ask though what it would take to give it a public home.
• 9:13
  SPEAKER_00: And if you're running one on LinkedIn already and then treating it as a brand awareness play, I'd also look into whether it's doing the work that you're not currently measuring.
• 9:22
  SPEAKER_00: Now, let's move on to our second story.
• 9:24
  SPEAKER_00: It is our current favorite drama, Google and Reddit.
• 9:27
  SPEAKER_00: Google's Jennifer Cuts told The Verge that Reddit gets no special preference in Google's ranking systems, and that Google's AI search features aren't built to display content from any specific site or platform.
• 9:39
  SPEAKER_00: Now, when I read this, my ears kind of perked up a little.
• 9:42
  SPEAKER_00: She also said there's a high and low quality content on forums like Reddit, the same as anywhere else, and that Google systems are designed to detect which is which, and that they've held results 99% spam-free for years.
• 9:56
  SPEAKER_00: Again, I really had to pay attention to this one.
• 10:00
  SPEAKER_00: On its own, that is a standard denial.
• 10:02
  SPEAKER_00: Google says versions of this about everything.
• 10:05
  SPEAKER_00: If I were reading this piece just to be reading it, I would probably move on and not think much more about it.
• 10:10
  SPEAKER_00: Except for two things.
• 10:11
  SPEAKER_00: First, the context it came out of.
• 10:14
  SPEAKER_00: That quote is from a Verge article asking whether Reddit can fend off a new wave of AI SCO spam.
• 10:21
  SPEAKER_00: So the question on the table was, is Reddit getting flooded by people gaming it specifically to land in AI answers?
• 10:27
  SPEAKER_00: And the Google answer is essentially, well, we're good at spam and Reddit isn't anything special.
• 10:33
  SPEAKER_00: Second, we have to think about what's happening to Reddit at the same time.
• 10:37
  SPEAKER_00: Glenn Gabe has been arguing that Reddit's traffic decline traces to the May Court update and the June spam update rather than to AI overviews, and he's pointed at Reddit's machine translated content in German and Italian as a major factor, noting that Steve Huffman's current explanation of those AI translations is a significant departure from how he described them when they launched.
• 11:00
  SPEAKER_00: Meanwhile, Reddit stock took a hit after the last earnings call with Huffman publicly saying AI overviews cannot replace Tim Blue links.
• 11:09
  SPEAKER_00: I want to be careful about the denial itself because I think there's a smidgen of a gap in it.
• 11:13
  SPEAKER_00: Gutz is saying Reddit gets no preference in Google's ranking systems.
• 11:17
  SPEAKER_00: That is not the same claim as Reddit doesn't appear disproportionately.
• 11:21
  SPEAKER_00: Practitioners have watched Reddit show up everywhere for a couple of years now, and a statement about how the systems are designed doesn't really settle what the systems produce.
• 11:31
  SPEAKER_00: Those two things can't be true at once.
• 11:33
  SPEAKER_00: But that's the SEO conversation, and it's not what why this really matters to us.
• 11:39
  SPEAKER_00: Here's why it matters for AI visibility though.
• 11:41
  SPEAKER_00: And I talked about this in the visibility report last week, which is my Friday deep dive.
• 11:46
  SPEAKER_00: It goes out every Friday straight to your email.
• 11:48
  SPEAKER_00: Link is in the show notes if you want on the list.
• 11:50
  SPEAKER_00: But I talked about Reddit as a third-party corroboration surface.
• 11:54
  SPEAKER_00: It's one of those places a model can go to find out whether anyone besides you says you're good at what you do.
• 11:59
  SPEAKER_00: It's part of why it's been such a popular play lately.
• 12:02
  SPEAKER_00: And every single thing that makes it valuable the volume, the appearance of independence, the fact that anyone can participate for free is exactly what makes it a target for manipulation.
• 12:13
  SPEAKER_00: And platforms respond to manipulation targets by, well, just discounting them.
• 12:17
  SPEAKER_00: So this week, Reddit is absorbing a core update, absorbing a spam update, fighting a wave of people gaming it for AI placement, publicly blaming the AI layer, and being told by Google that it isn't special.
• 12:30
  SPEAKER_00: If your brand's presence in AI answers leans hard on just one single platform, that platform's bad quarter is also your bad quarter.
• 12:40
  SPEAKER_00: And you won't have changed anything, you won't have done anything wrong, the ground just got revalued underneath you.
• 12:46
  SPEAKER_00: Mentions distributed across a lot of independent sources though survive a single platform having a bad quarter.
• 12:52
  SPEAKER_00: Mentions concentrated on one, though, survive absolutely nothing.
• 12:56
  SPEAKER_00: I'll say this too, since I'd be a hypocrite otherwise, I do use Reddit constantly.
• 13:01
  SPEAKER_00: I read it every week.
• 13:02
  SPEAKER_00: If you are connected with me on LinkedIn, it's where a lot of my found in AI LinkedIn edition newsletter questions come from.
• 13:09
  SPEAKER_00: And it's one of the best places to hear how people actually talk about a problem, especially relating to AI search visibility, which is what I'm usually looking at.
• 13:18
  SPEAKER_00: Reddit as a listening surface and Reddit as a citation strategy are two different uses though, and only one of them is really exposed to this risk.
• 13:26
  SPEAKER_00: So that's it for this week.
• 13:27
  SPEAKER_00: There were two stories.
• 13:29
  SPEAKER_00: Again, the week has been kind of quiet.
• 13:31
  SPEAKER_00: Microsoft shift a free tool for measuring which domains get selected as sources and defined the metric themselves.
• 13:37
  SPEAKER_00: Google said no domain gets preferred as a source.
• 13:40
  SPEAKER_00: One is out there building the measurement, one is out there denying the tilt.
• 13:44
  SPEAKER_00: Both are the industry arguing about the same question though.
• 13:47
  SPEAKER_00: How do the systems decide who to cite?
• 13:50
  SPEAKER_00: Nobody seems to agree.
• 13:52
  SPEAKER_00: The platforms don't agree with each other, the platforms don't agree with practitioners, and the tool vendors don't agree with anyone.
• 13:59
  SPEAKER_00: That's what an immature measurement layer looks like, and I think we're gonna be here for a while.
• 14:04
  SPEAKER_00: Which is the argument for not waiting.
• 14:07
  SPEAKER_00: If you sit out until the metrics settle, you're building authority from zero in a market where the people who started way earlier already have a citation history, and it's a lot faster to be found by a system that already knows who you are than to introduce yourself to one that doesn't.
• 14:22
  SPEAKER_00: So, two things to do this week.
• 14:24
  SPEAKER_00: One, find out if anyone at your company has Microsoft Clarity installed, and then find out who reads it.
• 14:30
  SPEAKER_00: Two, pull up your last 20 third-party mentions and then count how many different domains they're on.
• 14:38
  SPEAKER_00: If it's a short list, you have a concentration problem.
• 14:40
  SPEAKER_00: And now's a good week to learn that before anything else changes.
• 14:44
  SPEAKER_00: That's it for the week.
• 14:45
  SPEAKER_00: If you want to know where your brand actually stands before you change anything, the AI Search Visibility Audit is what I do.
• 14:51
  SPEAKER_00: The links are in the show notes, or you can just head over to CassieClarkmarketing.com.
• 14:55
  SPEAKER_00: I will see you Tuesday.
• 14:56
  SPEAKER_00: Until then, stay visible.