Google is testing exact and phrase match ads inside AI Mode. Ginny Marvin confirmed a small experiment that lets regular search campaigns serve text ads in AI Mode when intent is explicit and direct. On episode 20 of The GEO Show, recorded September 5, 2026, Paris Childress calls that the biggest story, then covers Similarweb and Baconi click cannibalization, Organic KPI citations outside the organic top 10, creator briefs, Amberos funding, Mention OS, product-page citations at 24.1%, entity-accuracy risk, crawler policy, and ChatGPT traffic up 48.4% while Bing fell 50%.
Google is testing exact and phrase match ads inside AI Mode. Ginny Marvin confirmed a small experiment that lets regular search campaigns with exact and phrase match serve text ads when intent is explicit and direct. Performance Max and AI Max remain the black-box tools for complex conversational intent.
Paris Childress recorded episode 20 of The GEO Show solo on September 5, 2026. He called the AI Mode ads test the biggest story of the day, then walked the redistribution layer around it: users versus clicks, citations outside the organic top 10, creator briefs, funding, product-page citations, entity accuracy, crawler policy, and uneven traffic growth across ChatGPT, Bing, LinkedIn, and Google.
Because they reconnect conventional paid-search keyword control to conversational search. Until now, Performance Max forced broad match, and AI Max removed keyword specification entirely. This experiment gives advertisers more direct control over return on ad spend inside AI Mode.
I think this is probably the biggest story of the day, although it might seem minor.
Paris's takeaway is to treat AI Mode as a combined earned plus paid visibility surface while keeping GEO and paid attribution separate.
Yes. Similarweb finds 95% of ChatGPT users also use Google even as generative AI visits grew 70% year over year. Baconi research found households with ChatGPT search access reduced traditional searches by about 9.4%. A randomized AI Mode experiment found external click-through rate fell 18.8 percentage points.
Audience reach, query volume, and clicks are different metrics. Stable Google usage and AI cannibalization can both be true. Separately, Organic KPI analyzed 1,208 citations across 160 commercial best-X queries and found only 53.6% of cited domains appeared in the same query's organic top 10. Indexed status matters more than top-10 obsession for citation eligibility.
Digiday reports Tinuity, DEPT, Jellyfish, and Ogilvy building AI discoverability into creator briefs, with Tinuity saying AI visibility rose from roughly 10% to about 25% of creator strategy considerations. Amberos raised $5.5 million (total funding $6.7 million) and is funding Creator Graph to map which creators and publishers influence AI recommendations. Mention OS launched an answer-engine visibility agent that monitors major engines daily, traces cited sources, and can draft site fixes.
10 Speed analyzed 7,387 citation appearances across 170 B2B buyer evaluation prompts on ChatGPT, Claude, Gemini, and Perplexity. Product pages generated 24.1% of citations, articles 17%, and Reddit plus YouTube plus forums only 4.2%. The study excludes Google AI Overviews and AI Mode. Paris said the Reddit-mention playbook is too simplistic. Product pages need specifics, integrations, compliance details, and comparison evidence.
Duane Forrester in Search Engine Journal describes entity-accuracy risk: when evidence is thin, AI may fill gaps with category norms, stale information, or better-documented competitors, and on-site audits may miss it. Crawler strategy must distinguish training bots from answer bots. Paris argues against accidentally blocking training-oriented bots if brands publish high information-gain content for GEO. Search Engine Land / Semrush estimates show ChatGPT.com US visits up 48.4% to 1.09 billion monthly visits, Bing down 50%, LinkedIn up 15.5%, and Google.com up 10.7%. Attention is redistributing unevenly across a fragmented discovery ecosystem.
00:27. Similarweb and Baconi: users stay on Google while searches and CTR fall.
03:17. Ginny Marvin: exact and phrase match ads tested in AI Mode. Paris's biggest story.
05:19. Creator briefs for AI citations; Tinuity AI visibility share roughly 10% to 25%.
11:05. 10 Speed: product pages at 24.1% of B2B buyer-stage citations.
16:31. ChatGPT +48.4%, Bing -50%, Google still +10.7%.
The throughline stays keyword control returning to AI Mode ads, with earned plus paid attribution kept separate while the discovery stack fragments.
Hi, everybody. Welcome to episode 20 of "The GEO Show." We are now 20 episodes in. That went by pretty fast. I'm your host, Paris Childress, and the show is brought to you by GEOforge: full self-driving for AI visibility. We've got some great stories to run through today on September 5th, so let's get right into it.
First up, AI isn't replacing Google users, but it is replacing searches and clicks. This one is kind of a big dbut it is coming from a very credible source, Similarweb. Similarweb finds that 95% of ChatGPT users also use Google even as generative AI visits grew 70% year over year. But separate Baconi research found households gaining ChatGPT search access reduced traditional searches by about 9.4%, while a randomized AI mode experiment found external click-through rate fell 18.8 percentage points.
All right, so 95% of ChatGPT users also using Google, that's no real surprise. I think what's more interesting is that, the reduced searches of 9.4%, I would've expected that to be even more. And of course, click-through rate falling almost 19 percentage points is a pretty big deal, and that's pretty clear evidence of the zero-click world that we live in now.
So I think that the two statements that Google usage is stable and AI is cannibalizing search can both be true. So the audience reach, query volume, and clicks are all different metrics Next up, nearly half of Google AI overview citations don't rank organically in the top 10. According to Organic KPI, who analyzed 1,208 citations across 160 commercial best X queries, and what they found was 77.6% of citations pointed to ranked lists, while only 53.6% of cited domains appeared in the same query's organic top 10.
YouTube and Reddit drove much of the user-generated content citation share. Okay, well, this is still another very strong data point that SEO rank and GEO citation eligibility are being decoupled, and we've now covered this theme numerous times here on the GEO show, and we just keep seeing more and more evidence coming up that no longer do you really need to try to get a page ranked in the top 10 to give yourself a good chance of winning for that page to win citations.
It really just needs to be indexed. top 10 ranking is no longer that critical All right, next story. Google is testing exact and phrase match ads inside AI mode, and I think this is probably the biggest story of the day, although it might seem minor. Google Ads liaison Ginny Marvin confirmed a new small experiment allowing regular search campaigns using exact and phrase match to serve text ads in AI mode when a user intent is explicit and direct.
I guess that Google will judge that based on the prompt or the query. So AI Max and Performance Max remain Google's tools for more complex conversational intent. So Google here is beginning to connect conventional paid search inventory directly to conversational search. So up until now, to give you all a little bit of background, if you wanted to do Performance Max in Google, you cannot specify a keyword match type.
You have to go with broad match, and you basically just have to let Google decide what keywords to match up and include in your, in your Performance Max campaigns. And the evolution of Performance Max is called AI Max, and that's the same thing. You basically cannot-- I think with AI Max, you can't even specify keywords at all.
but this is a very interesting move by Google because now they're testing more, the more granular controls of exact match and phrase match keywords for Google AI mode. And that's gonna give a lot more direct control of advertisers to try to engineer return on ad spend and to make Google Ads less of a black box when it comes to Performance Max and AI Max, especially within Google AI mode.
So I think we should be treating AI mode as a combined earned plus paid visibility surface while keeping attribution for GEO and paid placements separate Next story. Creator briefs are now being designed for AI citations. Another very interesting finding from Digiday. Digiday reports agencies including Tinuity, DEPT, Jellyfish, I know that one, and Ogilvy.
Okay, those agencies are incorporating AI discovery, AI discoverability into creator strategy. Tinuity says AI visibility has grown from roughly ten percent to about twenty-five percent of its creator strategy considerations. And DEPT is already briefing creators around exact product names, verifiable claims, chapter markers, and clean transcripts.
So what's happening here? What's happening here with the bigger picture? That influencer marketing, where you use creators as, basically to create your content, this is starting to merge in with GEO and AI search. Influencer marketing, digital PR also are both converging with GEO. So I think it's getting more and more important for marketers that wanna prioritize AI search and that are doing influencer marketing and creator marketing to really prioritize credible subject matter experts with deep domain authority, not just going after the mega influencers with huge follower counts.
Why? Because it's not as much about the pure reach of that creative content through the influencer, but it is about their subject matter expertise. So in this case, it's better to favor a smaller influencer with a smaller audience, but who can speak more credibly about the subject matter and who can provide higher information gain and inject more, information gain into that content All right, next up, a GEO/AI brand management startup has just raised $5.5 million.
Amberos announced a $5.5 million seed round, bringing their total funding to $6.7 million. Its platform monitors ChatGPT, Gemini, Claude, Perplexity, Grok, and DeepSeek, and it uses a brand knowledge graph covering 10,000 brands. Funding will also support Creator Graph, designed to map, to map which creators and publishers influence AI recommendations.
So this is very, an interesting nod to the last story. Creator Graph is gonna match... Oh, I'm sorry, map which creators and publishers are influencing AI recommendations. So the capital, so to speak here, is moving beyond AI visibility monitoring, and it's moving toward influence mapping and automated intervention.
So it'd be very interesting to see how this evolves, particularly with this Creator Graph product. I think it's very interesting, and it's pulling the world of influencer marketing closer and closer into GEO. Next up, Mention OS launches an autonomous answer-engine visibility agent Mention OS launched an answer-engine visibility agent that monitors ChatGPT, Gemini, Perplexity, and Google AI daily.
It identifies visibility changes, traces cited sources, and can draft fixes for publication to a brand's site. That last piece is the most interesting, of course, because I'm wondering, well, what kind of, what kind of fixes, drafting fixes for publication? but again, more broadly speaking, what we are seeing here is the continued trend that goes way beyond just monitoring, and it goes even beyond diagnosing, but also making specific recommendations and then executing on those recommendations.
So really, you can think of this as a closed-loop cycle. This is how we think of it at GEOforge. It does start with the measurement, and in fact, you really do need accurate measurement, high confidence measurement with low margin of error, but I won't go further into that. So once you get the measurement picture, then you need to make certain interventions.
You need to actually create content that will be published through your owned channels, like your website or maybe even your YouTube channel, but also earned channels where you can place that content, ideally high information gain content on third-party sites where you can leverage their authority because, AI is looking for that authority to basically validate your brand and your brand's claims.
And so once you have, you have measurement, then you have content intervention, and then you have distribution to owned and to earned, and then you have a remeasure cycle, and then you start that closed loop over again. And what you have then is proprietary data that informs the process through every one of these loops.
Every, round trip through these loops, you're gathering more data, the system is getting more intelligent, and the hope is that you can eventually build a flywheel from that. Okay, let's get right to the next story. Product pages dominate B2B buyer stage AI citations 10 Speed analyzed seven thousand three hundred and eighty-seven citation appearances across one hundred and seventy B2B buyer evaluation prompts spanning ChatGPT, Claude, Gemini, and Perplexity.
What they found was that product pages generated twenty-four point one percent of citations. Articles seventeen percent while Reddit, YouTube, and forums combined produced only four point two percent. All right, so this is a September third study, so this is post-Reddit decline, and this is across ChatGPT, Claude, Gemini, and Perplexity.
interestingly, it is not including Google AI Overviews or AI mode. So that might explain why YouTube has such a small relative presence in the citation share. But, um, the main theme of this headline is that product pages were number one. Product pages came up twenty-four percent of the time, and these are buyer evaluation prompts.
So this is saying that, you know, product pages are still very important, and product pages can legitimately make it into the citations. So the just get mentioned on Reddit GEO playbook, I think it's already come and gone, and it's far too simplistic. All right And really, so these product pages, I mean, you've gotta be very specific about, the product, what it does, how it integrates, its compliance, features, and comparison evidence because LLMs love, comparison content.
So I think that's one of the highest value add things to add to a product page is how its features and its value compares to other competitors in that same category. Next story. A new GEO risk. AI may describe your competitor while evidence about you is thin. Search Engine Journal's Duane Forrester describes the situation as such: "When evidence about a company is insufficient, AI systems may fill gaps using category norms, stale information, or better documented competitors.
Crucially, ordinary on-site content audits may not reveal the problem because it only appears in generated outputs." So GEO, as a reminder, most people think of GEO as all about visibility, but it is also very much about entity accuracy and brand accuracy and brand positioning So we always have to test, and to see not only is AI recommending our brand, but is AI saying accurate-- making accurate claims about our brand.
Next story, AI crawler strategy needs to distinguish training from answering. A September update on AI crawler behavior reports that publishers increasingly are blocking training-oriented bots while allowing search and user answer bots to crawl. Separately, Search Engine Journal notes that robots.txt remains voluntary while content delivery network, CDN, server, and web application firewall controls actually enforce crawler blocks.
So this theme has come up before, and it's becoming very important now to monitor AI bot crawling on your website and not to... In my opinion, not to accidentally block training-oriented bots because they are coming to gather information. They're seeking information gain so that they can update their training data sets.
And if you don't allow that to happen, you're basically blocking your high information gain content from getting that benefit. Even if you still allow search and user answer bots to crawl, that's fine because, um, these are all about the real-time search. So these two different types of bots are very different.
As I said, the training bots are there to update their training data. The search and user answer bots are there to support the real-time search that happens with the majority of prompts, where you have a prompt that gets converted into a query, a dozen or so queries, so-called query fan-outs, and then they go and search the web.
And that's why, of course, you need to be indexed in search results, not necessarily on the first page anymore, but you still need to be on the-- in the index. But getting back to the first point here, um, do not block training-oriented bots if you care about GEO and you're publishing high information gain content, because that's the whole point, is to feed these training-oriented bots with exactly what they're hungry for, your proprietary brand knowledge All right, up now we're on the last story of the day.
ChatGPT traffic rose 48% while Bing fell 50%. Ouch. reported on September 4th by Search Engine Land's analysis of Semrush traffic analytics estimates. ChatGPT.com US visits rose 48.4% year over year to 1.09 billion monthly visits, putting it 9th overall. Bing fell 50% while LinkedIn grew 15.5%. Google itself grew still 10.7%.
All right, so interesting. ChatGPT is a big winner here, and of course, we know that, um, while the major- I think the majority of ChatGPT users are using the web version, but still many people are using the, either the desktop app version, like I do, or the mobile app version. So 1.09 billion monthly visits is still not even representing the entire audience and the entire user base of ChatGPT.
it's sad news for Bing, but I'm not that surprised. LinkedIn's 15% growth kind of makes sense. And Google, meaning visits to google.com growing 10% at this stage is pretty surprising to me. So, this is, this is much more complex than just saying AI is replacing Google. really what's happening is that attention is being redistributed unevenly across a more fragmented discovery ecosystem.
So that actually makes things more interesting for marketers, in my opinion, because now our visibility strategies and our budgets can expand from just Google SEO to Google plus AI engines, plus LinkedIn, plus earned source ecosystems, um, rather than just shifting everything from one channel to another.
think about it more as an uneven redistribution. All right, everybody. That's it for today. we had a great episode 20. Congratulations to The GEO Show on breaking 20 episodes. We're gonna keep going here, and we'll see you all in the next one.
