ChatGPT is testing ads that open branded business agents inside the chat instead of external landing pages, with Wayfair in an early Digiday-reported pilot. On episode 32 of The GEO Show, Paris Childress covers Profound's global Estée Lauder GEO deployment, Cloudflare's September 15 AI crawler defaults, Supertab Connect edge bot intelligence, Indexa's Turkey launch, a comparison page that jumped from 4 to 49 citations, Stripe mentions versus Airwallex owned citations, AirOps citation-to-revenue research, Ahrefs Brand Radar at roughly 457 million prompts, and TrustPoint Exposure's five-agent GEO suite.
ChatGPT is testing ads that open branded business agents inside the conversation instead of sending users to external landing pages. On episode 32 of The GEO Show, Paris Childress explains Digiday's September 14 report that OpenAI showed select advertisers this format, with Wayfair in the pilot, and frames it as early commercial experimentation that matches the prediction that AI ads will be in-conversation agents rather than banner cost-per-click placements.
Around that throughline he covers Profound's global Estée Lauder GEO deployment, Cloudflare's September 15 AI crawler defaults, Supertab Connect edge bot intelligence, Indexa's AI brand visibility launch from Turkey, a Search Engine Land comparison-page experiment that jumped from 4 to 49 citations, Stripe mentions versus Airwallex owned-source citations, AirOps citation-to-revenue research, Ahrefs Brand Radar at roughly 457 million prompts, and TrustPoint Exposure's five-agent GEO suite.
Because staying in chat is a better product experience than ripping the user into a cold landing page. Digiday reported that sponsored placement can open a branded business agent directly inside ChatGPT. Wayfair confirmed involvement. OpenAI has not broadly launched the format, so it remains an early experiment. Paris argues banner ads and classic CPC will not be the durable model: advertiser agents should enter the conversation at the right moment and let users transact where they already are.
I think this is the future of ChatGPT ads, and it's the future of AI ads in general and other LLMs that I think will follow.
Brand marketing. The Estée Lauder Companies announced a global strategic partnership with Profound on September 14 to deploy across the full portfolio and measure representation and recommendation across generative AI platforms. Paris congratulates Profound as the current category leader and stresses the larger signal: GEO is about how brands are represented, presented, recommended, and positioned inside AI responses, not about ranking for a click and converting at the bottom of the funnel. Experimental budgets are becoming global enterprise infrastructure.
Check preferences now. Effective September 15, new defaults allow AI search crawlers while blocking training and agent activity on ad-supported pages for newly onboarded domains, and mixed-purpose crawlers inherit the restrictive policy. Free customers without explicit preferences are affected. Paris says unless you are a major news publisher protecting proprietary content, allow these bots so you keep surface area for inclusion, mentions, and citations. Supertab Connect, launched the day before, adds an edge layer that classifies retrieval, agents, and training traffic for visibility rather than blocking.
In one Search Engine Land structured experiment spanning 775 citation events, changing an owned comparison page to include leading competitors increased citations from 4 to 49 (more than 12×) without changing domain, author, or targeted keyword. Roughly half of cited sources in the wider experiment stopped being cited within 30 days. Paris connects that to battle-card pages and suggests multi-competitor comparisons may outperform simple A-versus-B pages when written objectively.
No. In 5W's payment study (297 answers across 24 prompts, three runs, four engines), Stripe appeared in 80% of answers with only 13% owned-domain citations, while Airwallex appeared in 12% with 44.3% owned citations. Paris argues mentions are the real win and citations are a consolation prize with abysmal click-through, way below 1%. AirOps research also surprised him: AI citations correlated more with non-branded than branded organic, possibly because people validate LLM answers with category searches on Google. Ahrefs Brand Radar's ~457 million prompts show scale without custom prompt discovery, and TrustPoint Exposure shows PR agencies packaging GEO as autonomous agents ahead of an October 1 launch.
00:29. Estée Lauder × Profound: first prestige beauty GEO deploy at this global scale.
02:23. ChatGPT ads open branded business agents inside chat; Wayfair in the pilot.
04:36. Cloudflare AI crawler defaults effective September 15; free accounts must check prefs.
09:48. Owned comparison page: 4→49 citations (12×) by adding leading competitors.
11:38. Stripe 80% mentions / 13% owned citations vs Airwallex 12% / 44.3%.
The throughline is surfaces: in-conversation branded agents, enterprise brand marketing for GEO, and crawl access decisions now decide whether brands show up in AI answers.
Hi, everybody. Welcome back to another episode of The GEO Show, brought to you by GEOforge: Full self-driving for AI visibility. Today, we are on episode 32. I'm your host, Paris Childress, and we got some great stories to get into, so let's kick it off. First up is Profound lands a global Estée Lauder deployment. The Estée Lauder Companies announced a global strategic partnership with Profound on September 14th.
The company says it will deploy Profound across its full portfolio and become the first prestige beauty company to operationalize GEO globally at this scale. The stated goal includes measuring how brands are represented and recommended across generative AI platforms and identifying optimization opportunities. This is pretty major, and congrats to Profound. I probably would expect a big brand to go with Profound because right now they are the leader, but this is about something much bigger.
A major company here, enterprise company, is partnering with a GEO platform provider to measure their entire brand portfolio, and this is not performance marketing anymore. This is brand marketing, and I think that is actually where GEO diverges most acutely from legacy SEO, is that it's brand marketing. It's fundamentally about how your brand is represented, presented, recommended, positioned inside of AI responses.
It's not about performance, meaning ranking, getting the click, and then winning that conversion or that sale at the bottom of the funnel. So it's moving also from experimental marketing budgets into global enterprise infrastructure. So this is great news for the category overall, and I would expect more announcements like this to come. Next up, ChatGPT tests ads that open branded AI agents instead of websites. ChatGPT tests ads that open branded AI agents instead of websites. This is also fascinating and pretty major.
Digiday reported on September 14th that OpenAI has shown select advertisers a new format in which sponsored placement opens a branded business agent directly inside ChatGPT rather than sending the user to an external landing page. Wayfair confirmed involvement in the sponsored agent pilot. OpenAI has not broadly launched the format, so this should be treated as an early commercial experiment rather than a general product release.
Very interesting and very timely because Simon and I just debated this last Friday on our LinkedIn Live episode, which is, what is the future of ChatGPT ads going to look like? I don't think people are going to respond well to banner ads placed somewhere on the screen. I don't think it's gonna be a cost-per-click model either.
So this is actually what we predicted, which is that agents acting on behalf of the advertiser, the brand advertiser, will step into the conversation at the appropriate time and then allow the user to transact effectively inside of the LLM platform. In this case, inside of ChatGPT.
So rather than going, clicking away from ChatGPT into a website and being ripped out of that chat experience that you want, and then going into some website for the first time that you have to understand and navigate and react to a different environment, that's not a great experience.
It's a much better experience to continue speaking with an agent who you trust inside of the chat platform where you already are, and then have a smooth transition right into the shopping experience right there. I think this is the future of ChatGPT ads, and it's the future of AI ads in general and other LLMs that I think will follow. So we're gonna watch this very closely.
Next story is a final reminder that today, effective September 15th, Cloudflare's new AI crawler defaults take effect. This is a reminder. We've been repeating it several times. The new defaults allow AI search crawlers while blocking training and agent activity on ad-supported pages for newly onboarded domains. Mixed-purpose crawlers inherit the more restrictive applicable policy. So existing free customers who had not explicitly set preferences are also affected by the new defaults.
So many people who are on a free Cloudflare plan, you all need to pay attention to this because if you do nothing, then it very well could be that training agents are going to be blocked by default or mixed agents that combine search and training, mixed purpose agents will also be blocked by default.
And unless you're a major news media, and you don't want your proprietary content to be crawled and you don't want to be included in LLM responses, I think that you should allow all these bots to crawl because it's only gonna improve and increase your surface area for getting included, mentioned, and cited in LLM responses. So if you have a Cloudflare account, free or paid, go in there and check those settings today. It's the last chance.
Next up, Supertab launches an AI traffic intelligence layer at the edge. Supertab launched Supertab Connect on September fourteenth. It classifies automated traffic by operator, activity category, and requested URL, and it can run through Cloudflare, Fastly, CloudFront, WordPress, or an SDK. The product focuses on visibility and reporting rather than blocking and distinguishes activities such as retrieval, agents, and training.
So just on the heels of that last Cloudflare update, we have a new tool called Supertab Connect, and it's also analyzing activity from AI bots. And this is becoming really critical infrastructure for SEO and for AI visibility. Understanding which bots are crawling, how they are crawling, how often are they crawling, which pages are they crawling. This crawling activity is the prerequisite for visibility.
You can't be cited in an LLM engine unless you are first crawled and indexed in many cases, and the LLMs are now, as we've seen, building their own indexes. So AI bot intelligence is emerging as a separate layer, a software layer between infrastructure and GEO analytics, and we need to start watching this more carefully. Next up, Indexa enters the AI brand visibility market.
The first big player out of Turkey, Istanbul-based Indexa announced an AI brand visibility platform designed to track how companies appear across AI systems, compare competitors, surface gaps, and monitor signals, including brand mentions, recommendations, citations, sentiment, and AI share of voice. So the first big player that we've seen coming out of Turkey, I checked out Indexa, and it does seem to be another tool that is attempting to merge SEO and GEO rather than a pure play GEO tool.
And I do think that as the category matures and as the category is still taking shape I think that the tools that are still carrying a lot of legacy SEO features and trying to merge that into a larger feature set with GEO features as well, I think that's a losing proposition. I think that that is a tool that is trying to be all things to all people.
It's trying to straddle the old SEO world and the new GEO world, where I believe that actually the best strategy to win in GEO and to at least maintain and maximize whatever traffic potential is left from SEO is to put GEO first and to have a GEO-first strategy. And a feature set if you're in the tool space, the feature set should focus on GEO.
When I opened Indexa, I saw that the first step of the recommended workflow is to conduct keyword research, so that's clearly a legacy SEO tactic. And I think that, in my opinion, there's really no place for keyword research in a GEO-first strategy. On to the next story. One GEO experiment moved the page from four to forty-nine citations. Search Engine Land reported two structured GEO experiments spanning seven hundred and seventy-five citation events.
In one test, changing an owned comparison page to include leading competitors increased its citations from four to forty-nine, which is a more than twelve x increase. Without changing the domain, the author, or the targeted keyword. Roughly half of the cited sources in the wider experiment stopped being cited within thirty days. So many of us out there are building comparison pages with competitors. We've done probably thirty of them. We call them battle card pages. It's GEOforge versus competitor X.
And what this is showing is that when you increase the number of competitors here, and I presume that you still are doing this in an objective way, changing that comparison page to include leading competitors increases citations by an order of magnitude, in this case twelve x.
So maybe what that means is instead of having your us versus them or A versus B, maybe you have competitor pages that are comparing yourself to a larger group of competitors and maybe getting a little bit more strategic about how you're different within your category to several of the leading tools as opposed to just one of the tools. But it is interesting. It has given me some food for thought about how to also experiment more with our competitor battle card pages.
Next up, Stripe dominates mentions, but Airwallex dominates owned source citations. 5W analyzed two hundred and ninety-seven AI answers across twenty-four payment industry prompts, repeated three times, not really enough, but three times across ChatGPT, Gemini, Claude, and Perplexity. Stripe appeared in eighty percent of the answers, but it had its own domain cited in only thirteen percent. Inversely, Airwallex appeared only 12%, but its domain was cited in 44.3% of retrieval evidence answers. An interesting play here.
Stripe is clearly the larger, more well-known brand, so it's no surprise that they are winning on mentions for these prompts. But Airwallex is winning on citations, and as we've proven before in other stories in previous episodes mentions are not the same as citations. Mentions, in my opinion, are much more valuable than citations, and they're not even correlated. Mentions is really what you want.
That is the ultimate definition of winning in AI visibility is having LLMs mention your brand and recommend you in its answer, and the consolation prize, I think, is citations, which means that you're cited as a source that AI referred to and pulled into its synthesized answer. But as we've seen, and I really believe, that most people aren't really paying attention to the citations. The click-through rate is abysmal. It's way, way below one percent.
So they're not really looking at citations too much. They're not paying attention. They're definitely not clicking through, and where you really win is on the mentions. And I think what this is showing us between Stripe and Airwallex is Stripe is winning because of its off-site activity, because its brand is probably mentioned in far more places around the web than Airwallex. Yet Airwallex is winning, or at least they have more citation links appearing probably because they're doing more dedicated GEO pages.
They're focusing more on owned content, publishing more pages that are optimized for GEO in terms of formatting and answering questions and publishing that on their website as opposed to Stripe. I'm not even sure if Stripe really is paying attention to GEO because they're so dominant anyway. All right, onto the next story. AirOps sharpens its citation-to-revenue positioning. AirOps published a September fourteenth analysis that synthesizes six studies across more than eighty-seven domains.
It reports that AI citations correlate more strongly with non-branded than branded organic traffic. That is a big surprise to me. They cite research suggesting that AI-referred visitors convert substantially better than traditional organic visitors. That we've covered before. And it argues that citation effects can appear quickly. That's also interesting. let's unpack this one by one first we have branded versus non-branded.
This is the big surprise to me because I always expected that as you get more AI citations, and generally as your AI presence and your AI visibility grows within LLM responses, more people are going to search for your brand. So I would expect a stronger correlation with branded search and branded organic traffic as opposed to non-branded. This study says the opposite, so this is a big surprise to me. It is a stronger correlation with non-branded than branded.
So apparently, people are going from LLM responses where they might see your brand, and instead of searching for your brand, they're searching for a non-branded keyword, maybe a category keyword. And it could be that they're doing that not to navigate to your website because they're already convinced and are ready to convert, but rather they're going to Google and trying to validate the answer that they found on ChatGPT.
And that kinda makes sense actually, because some people are still looking to use search as a backup or a second validation step after they start from an LLM, and I think this is common behavior. So I'm starting to now get more convinced why this makes sense. More non-branded organic traffic than branded organic. The next point here is that AI visitors convert substantially better than traditional organic visitors.
That's no big surprise because LLMs have educated them and brought them much further down the funnel than they would typically start on a search. And we've seen that across the board, much higher conversion rates from AI-referred visitors. And the last is that citations can appear quickly.
And I agree with this too, because there is a freshness factor, and there have been studies shown that citations or pages that have been created in the last six months have an outsized share of overall citations. So that means even if you're losing or if you feel like it's too late to get started, it's not. Because if you start creating content, you can win and you can overtake or leapfrog your competitors because your pages are fresher than theirs.
And by the way, the same thing is very true with video, with YouTube videos. Next story. Ahrefs' Brand Radar is operating at roughly four hundred and fifty-seven million tracked prompts. Ahrefs Brand Radar currently reports roughly four hundred and fifty-seven million search-backed monthly prompts across Google AI Overviews, AI Mode, ChatGPT, Perplexity, Gemini, and Copilot, with Claude supported for custom prompts. Ahrefs also connects AI visibility to web analytics and AI bot traffic. All right.
It's very hard to compete with the sheer scale of what Ahrefs has. Four hundred and fifty-seven million prompts most likely contains your category, but you don't really get to specify exactly which prompts. So what Ahrefs doesn't do is allow you to build your own custom prompt set and then track AI visibility for those custom prompts.
You go in there and say that this is basically this is my industry, this is my category, and it's gonna pull its own prompts from its gargantuan dataset, and it's gonna try to show you your AI visibility in your category from a sample, a large sample of prompts that they have classified in your industry. I don't really think this is gonna stick.
Ahrefs is certainly one of the kings of legacy SEO here but I just don't think this is the right way to do prompt tracking. And yes, I know that there are advantages to having connectivity to web analytics. I think what they're doing with AI bot traffic reporting is also interesting, but of course, we can get that already in Cloudflare.
So Ahrefs has a massive data advantage, but I do think the lack of having a feature to allow custom prompt list generation and real prompt discovery is gonna mean that Ahrefs is not gonna be one of the leading GEO tools six to twelve months from now. Last story. An agency is packaging optimization as five autonomous agents. TrustPoint Exposure announced a five-agent suite scheduled for public launch on October first, covering lead generation, trust authority, AI recommendations, and deal conversion.
The firm describes itself as the first certified PR agency for AI search optimization. That designation is a company claim rather than an independently established category credential. The first AI-certified PR agency. Okay, who certified them? We don't know. But a PR agency, as I've said before, I think GEO is a natural place for PR agencies to play in due to the fact that it is mostly dominated by off-site work, off-site citation building. And a lot of that is PR.
It's classic PR outreach to get your brand mentioned in places, to get your brand included in stories, and generally to do PR work. So it is kind of interesting here that TrustPoint Exposure is positioning these as agents. I think that is just generally a good way to market a service now is to discuss that you have agents.
And of course, they have the same sort of service flow where you start with measurement, you move towards action, and then you monitor the results of those actions and ultimately try to convert your brand's customers at the end of that process. So we'll see what happens with TrustPoint Exposure, and I do expect many PR agencies to be making that transition into GEO. All right. That's it for today. Thank you all for listening, and we'll see you on the next one.
