HubSpot put GEO and answer-engine visibility workflows inside ChatGPT and Claude via MCP, making the dashboard increasingly optional. On episode 26 of The GEO Show, Paris Childress covers Writesonic's ChatGPT-6 Astra shift toward brand-owned pages (71.1% first-party citation rate), a brief Google AI Mode citation bug, search ads entering AI Mode on direct intent, Brand Ghost's 12.6% cross-engine source overlap, Highwire Acro Answers in PR budgets, Matter's review-to-AI-discovery stack, Soupa Intent's ChatGPT ads library, Jessica Bowman on ops-driven AI narratives, and Meta buying Stilla.ai for agentic commerce in messaging apps.
HubSpot put GEO and answer-engine visibility workflows inside ChatGPT and Claude through its MCP connector. On episode 26 of The GEO Show, Paris Childress argues the dashboard is becoming optional, then walks ten stories spanning model citation shifts, Google AI Mode bugs and ads, cross-engine fragmentation, PR and reputation tools, paid AI libraries, operations-driven narratives, and agentic commerce.
The throughline is callable AI search intelligence. Visibility, citations, competitor share of voice, tracked prompts, recommendations, and content generation increasingly need to run where agents already work, not only inside a dedicated GEO UI.
HubSpot updated documentation on September 9 so users can retrieve visibility and citations, manage tracked prompts, review recommendations, and trigger content generation from ChatGPT or Claude via the existing MCP connector. Paris calls it the most important story of the day because the MCP layer keeps winning and the dashboard becomes optional.
The dashboard is becoming optional, and AI search intelligence increasingly needs to be callable and actionable by agents.
Writesonic tested fifty prompts across sixteen categories and found Astra averaging 6.1 fan-out searches versus 10.3 for ChatGPT-5.6 Sole, reading 50 web results versus 82, directing 83% of search activity at brand-owned pages, and posting a 71.1% first-party citation rate. Shallower fan-out plus branded preference can reset GEO share of voice when models update.
No. Brand Ghost's September 8 observatory run reports 12.6% cross-engine source overlap, 44.7% of comparable recommendation sets with zero brand overlap, and 40% of dated citations pointing to content more than a year old across ChatGPT, Claude, Gemini, and Perplexity. Blended AI visibility scores conceal engine-level source and recommendation behavior.
After Gemini 3.8 Flash entered AI Mode on September 2, some broad queries briefly returned fewer citations or none until Google fixed behavior Robbie Stein called unintended. Separately, Ginny Marvin confirmed a small experiment letting exact and phrase match keywords serve text ads in AI Mode on explicit direct intent. Split organic answer visibility from paid AI placement visibility.
Highwire's Acro Answers tracks ChatGPT, Gemini, Claude, and Perplexity for PR and corporate communications teams. Matter reframes review capture and reputation monitoring as AI discovery infrastructure. Jessica Bowman tells Search Engine Land that AI can ingest complaints, release notes, earnings calls, and employee reviews, so narrative remediation may leave marketing. Soupa Intent's ChatGPT ads library and Meta's Stilla.ai acquisition extend the map into paid AI competitive intel and agents that transact inside WhatsApp, Messenger, and Instagram.
00:25. Writesonic: Astra at 6.1 fan-outs and 71.1% first-party citation rate.
01:28. Google AI Mode citation bug after Gemini Flash swap.
02:28. Exact/phrase match ads eligible in AI Mode on direct intent.
03:28. Brand Ghost: 12.6% cross-engine source overlap.
06:13. HubSpot: GEO workflows callable inside ChatGPT and Claude.
13:50. Meta buys Stilla.ai for Business Agent commerce in chat.
The throughline is agent-callable GEO: when workflows live inside ChatGPT and Claude, the product is no longer only a dashboard.
Paris: Hi, everybody. Welcome to episode 26 of "The GEO Show," brought to you by GeoForge, full self-driving for AI visibility. I'm your host, Paris Childress, and I am re-recording this episode because the first version didn't capture the screen sharing properly, so I'm doing it again, and I'm doing it right. So let's kick off today's first story, which is about GPT Astra.
Paris: GPT-6 Astra searches less and goes straight to brand-owned pages. This is research from Writesonic, who tested 50 prompts across 16 categories and found that ChatGPT-6 Astra averaged 6.1 fanout searches per answer versus 10.3 for GPT-5.6 Sol while reading 50 web results versus 82 from Sol. It also reports that 83% of Astra's search activity targeted brand-owned pages and measured a 71.1% first-party citation rate.
Paris: The test used one run per prompt, which we know is not really enough, and one US Plus account, so the findings are directional rather than universal. All right.
Paris: What are we seeing? ChatGPT-6 Astra is a really big deal. People are raving about it, but like with all new models, it can really disrupt your share-of-voice measurement.
Paris: It can really provide more instability because what we're seeing here, at least so far with GPT-6, is we are seeing fewer fanout searches, so it's going, a little bit more shallow when it comes to real-time search, and it's also favoring much more heavily branded websites for its citations. So 83% of the search activity was targeting brand-owned pages, and 71% of the citations were first party, meaning from the brands themselves, not third-party editorial or UGC.
Paris: So keep that in mind that your share-of-voice tracking should always be also monitoring changes to the underlying models themselves, especially for ChatGPT. Let's move on to the next story. A Google model swap briefly removed AI mode citations. After Gemini 3.8 Flash entered AI mode on September second, analysts reported some broad queries returning sharply fewer citations or none at all.
Paris: Google Search VP Robbie Stein described the behavior as not working as intended, and source links subsequently began returning after a fix. So bugs can happen even at a large scale, like in Google. in this case, it was Google AI mode. but again, this is another case of a model update either disrupting completely or having very different behavior in terms of citations than the prior model.
Paris: So we always need to make annotations in our share of voice trend reports every time a model gets updated. We just saw evidence with ChatGPT Astra and here in Gemini 3.8 Flash, both of which were launched last week. On to the next story. Standard search ads can now enter Google AI mode for direct intent.
Paris: Google Ads liaison Ginny Marvin said, "Exact and phrase-match keywords are eligible to serve text ads in AI mode when Google detects explicit and direct user intent. Google describes it as a small experiment. Complex conversational intent remains the domain of AI Max and Performance Max. No public scale or performance data exists."
Paris: So we reported something similar on this before, which is that Google was experimenting with exact and phrase-match keywords inside of AI Max and Performance Max, which previously has not been allowed. And now this is kind of generally the same story because it's saying that a Google Ads user can use exact and phrase-match keywords and make them eligible to serve text ads in Google AI mode.
Paris: Up until now, Google AI mode was reserved only for AI Max and Performance Max, which had been broad match keywords only. So this is a little bit on the paid side, but I think more and more you're gonna see, just like with PPC and SEO, you're gonna see a split Previously, your search visibility always needed to be reported separately between what's coming in organically and what's coming in through paid.
Paris: And the same thing is gonna be true here with GEO. On to the next story. Fresh cross-engine data still shows no single AI search reality. This is a story from Brand Ghost Observatory, and they have confirmed on their September eighth observatory run that twelve point six percent cross-engine source overlap, which is very small, forty-four point seven percent of comparable recommendation sets that have zero brand overlap, so that's actually a large number, almost half having no brand overlap at all, and forty percent of dated citations pointing to content more than a year old, and this is across ChatGPT, Claude, Gemini, and Perplexity.
Paris: I don't know how large this data set is, but it does provide yet more evidence of cross-engine inconsistencies, basically. So across ChatGPT, Claude, Gemini, and Perplexity, we only saw a cross-engine overlap of twelve percent when it came to AI visibility. So it is very, very important to silo your share of voice and AI visibility reporting per LLM engine, per AI engine, because they differ so much. Next story.
Paris: Highwire launches Acro Answers, an example of a large PR agency now shipping AI visibility software. Communications agency Highwire launched Acro Answers yesterday. It tracks ChatGPT, Gemini, Claude, and Perplexity and organizes measurement around brand visibility, prevalence, perception, and topic adjacency.
Paris: Those are all very interesting. Highwire says it tracks two hundred and sixty-seven plus prompts and thirty thousand AI responses per brand per month. That comes out to about, I think, if my math is correct, a little more than one hundred runs per prompt. So two hundred and sixty-seven prompts, thirty thousand AI responses, and that's per month.
Paris: If they do it once per month, that's ten runs done per month. I'm sorry. No, it's actually one hundred runs per month. So maybe if they do that weekly, it's probably twenty-five runs. That, to me, is statistical relevance. I give that a big thumbs up. So good job there for Highwire. but the main theme of this news report is that a major PR agency is launching their own AI visibility software tool. So GEO is now, bleeding into other domains within marketing beyond performance marketing, which is where SEO always lived and was always pretty much siloed. PR definitely influences GEO, and so GEO is starting to encroach into PR, but not only PR, I would say heavily into brand marketing as well and going after those budgets.
Paris: Next story. HubSpot moves GEO work directly into ChatGPT and Claude. This is our top story of the day by far. Many of you know that HubSpot GEO is a new product that they've launched probably two months ago, and it is HubSpot's major play for GEO. As of yesterday, they've updated their documentation to support GEO workflows through its connector, through its Claude MCP connector, and I suppose it's now Claude and ChatGPT connector.
Paris: Users can retrieve visibility, citations, and competitor share of voice. They can create and manage tracked prompts. They can review recommendations, and most interestingly, they can trigger content generation without leaving ChatGPT or Claude. So one of the most powerful MCP connectors, Claude connectors in all of marketing has been HubSpot. And now that connector has extended into HubSpot's newest tool, HubSpot GEO.
Paris: Very, very interesting, and it's further proof here that we don't really need to go into these tools any longer, as long as there is an MCP connector. Claude and other, AI agents essentially are gonna be a middle layer between us human users and our MarTech stack. So the dashboard, let's say a dashboard sitting in HubSpot GEO is kind of optional really.
Paris: AI search intelligence is increasingly callable and actionable by agents. All right, next story. This next story matters because it's about Matter. Matter is a company that turns review management into an AI visibility service. Okay. PR agency Matter launched Review Visibility Management on September 9th, explicitly connecting review capture, reputation monitoring, local profiles, search visibility, and AI discovery.
Paris: Matter's causal claims about reviews and AI recommendations are vendor claims rather than independently demonstrated engine weights. All right, what is this all about? This is definitely a local GEO play, and it's also very heavily aimed at the reputation monitoring aspect of GEO.
Paris: This story was reported by Yahoo Finance, and what we're seeing is that reputation management, which has long been a discipline within marketing and PR, is now through this tool, through this new tool from Matter, it's being reframed as GEO infrastructure. And I think this puts the larger picture into focus, which is that GEO right now, the practitioners are mostly focused on visibility.
Paris: Is my brand showing up? Is it being mentioned? Is it being cited? But after that comes the question, if it is being mentioned and cited, what are they saying? Are they saying accurate things? What is the sentiment analysis? How is this affecting my reputation management? Because the reputations from review sites and other places are gonna get pulled in here from local profiles. So this reputation management is also now another area where GEO is encroaching. I would add that to the list: PR, reputation management, brand marketing. So something to keep in mind as this space evolves. Okay, the next story is from SupaIntent.
Paris: Paid AI visibility is becoming measurable competitive intelligence. So SupaIntent's public ChatGPT ads library shows three thousand one hundred and fifty-seven unique creatives and almost eight thousand recorded appearances with its feed refreshed on September tenth.
Paris: Its Ads Radar product tracks advertiser, message, landing page, prompt, position, and daily changes across ChatGPT and Google AI mode. Really interesting. SupaIntent explicitly cautions that an appearance is not an impression, and it provides no spend, targeting, conversion, or performance data. I think this is not the first tool, but it's one of the early third-party tools that is going out and taking snapshots of ChatGPT ads all around the place and building a library, for competitive intelligence.
Paris: And I do think that OpenAI will need to, at some point, launch an ads transparency center the way that Google has to make all the advertisements of every brand in its platform, every advertiser should be fully transparent. It should be made public. And of course, this is a phenomenal research tool. You can go in and research your competitors, or you can just generally see who are, what are some of the creative approaches that some of the top brands are taking in ChatGPT ads.
Paris: And you can see their ad copy, their creative. You can check out their landing pages. So it's probably a great tool, and we are all waiting eagerly for OpenAI to launch its own official tool for ads transparency. All right, next story. Back to AI reputation. AI visibility is becoming an operations problem, not just a content problem.
Paris: Enterprise SEO veteran Jessica Bowman told Search Engine Land that AI systems can incorporate customer complaints, release notes, earnings calls, vendor case studies, and employee reviews when evaluating companies. She describes AI as a sophisticated buyer advisor and argues that marketing can hit a ceiling when the underlying problem is operational.
Paris: This is a great, great story, and it really puts a spotlight on the fact that AI visibility and GEO also reaches far beyond just marketing. It's more than just your website that's getting crawled and then that information is being used or other third-party websites. these AI crawlers are going really, really deep. They're going into product release notes, earnings calls, vendor case studies that could be published all around the web. And really a lot of the times if you have an AI reputation problem that emerges, it's not necessarily a marketing problem. maybe marketing is gonna be expected to try to fix the problem, but in many cases, the narrative diagnostics should really identify which source class is driving this claim, and it should route the remediation to the appropriate department.
Paris: So it could be marketing, but it could be product, it could be customer success, HR, legal, or some other function, some other operational function. And so I think it is incumbent on us as early GEO practitioners to act accordingly when we have an AI reputation problem that emerges for our brand. We should make sure that the underlying cause of this problem gets addressed.
Paris: It might be that we need to go to the legal team and clear something up or maybe the HR team if there are some, well, let's say some problematic employee reviews. So it really, potentially touches all aspects of the business. All right, now we're on our last story of the day. Meta buys Stilla.ai to accelerate business agents that transact in chat.
Paris: All right, this is right on the heels of the Muse update, which I've been checking out, and I'll get back to that in a second. But this story is about Meta's acquisition of Stilla, which was confirmed yesterday, September 9th. Stilla is joining Meta. Axios reports that Meta has acquired the Swedish startup to strengthen Meta Business Agent, which facilitates commerce through WhatsApp, Messenger, and Instagram.
Paris: Meta says that more than one million businesses already use Business Agent. And one of the other major announcements, which was from the last, I believe, seven days, was the launch of Meta's newest AI agent called Muse, M-U-S-E. And Muse is really kind of like your personal assistant for life. It lives inside of WhatsApp, Messenger, Instagram, wherever you want it to be in any of these chat applications.
Paris: And it can do pretty much anything you ask it to do, and that includes help you to shop more smartly and to transact. So I think this purchase of Stilla.ai is directly connected with the Muse AI assistant and with its Meta Business Agent product, and I can see this all coming together. And tomorrow, Simon and I are gonna discuss this more deeply, which is the future of advertising and shopping in the age of AI.
Paris: And what we're gonna discuss is, are we ever gonna really need to leave our AI assistants anymore, these surfaces? do we actually need to click away from those surfaces into e-commerce websites or other types of websites where we might book things or order things? when we can do it right there.
Paris: So this is really a fascinating look into the future, and put that in the perspective of GEO. It's basically, first, you do need to be visible, and then you need to be recommended favorably with the right kind of sentiment and accurate information. And then beyond that, you've got to facilitate agents' ability to order and transact on behalf of their users.
Paris: And that's gonna be the last real surface area that I think a lot of people aren't really thinking of yet. But the first way to start thinking about that is to think of your website now primarily as a place where AI bots will outnumber human visitors, and the experience of those AI bots has to be seamless.
Paris: So right now, most people are thinking about that their websites need to be crawlable by AI bots, but that's just the very beginning of it. In the later stages, it's gonna be these AI bots are gonna need to be able to click through, navigate, just as a human would do, filter, use any kind of advanced features, and then actually transact or fill out forms or make orders and complete transactions.
Paris: So it's an exciting future. All right. That's all we've got for episode twenty-six. Thanks for listening, and see you all in the next one.
