OpenAI sponsored agents turn a ChatGPT ad click into a separate, clearly labeled conversation with an advertiser AI representative (limited alpha), reframing paid AI discovery as organic plus paid plus branded agent tracking. On episode 35 of The GEO Show, Paris Childress covers that shift alongside HubSpot as OpenAI's first CRM integration for ChatGPT Ads, Profound Aim's marketing OS ambition after the $180M Series D, Peec Brand Perception, Google Merchant Center AI share of voice, Lululemon conversational attributes in AI Mode, Ahrefs France CTR erosion, Cited Index market saturation, Miraqo repeated-run sampling, and GeoJobs as a career path.
OpenAI just turned ChatGPT ads into branded AI agents. On episode 35 of The GEO Show, Paris Childress opens with Profound Aim's marketing OS ambition after the $180M Series D, then Peec Brand Perception, before the locked throughline: sponsored agents that let a user click a ChatGPT ad and start a separate, clearly labeled conversation with an advertiser AI rep. HubSpot closes the CRM attribution loop so ChatGPT Ads can optimize on contacts, deals, lifecycle, and form signals. The rest of the briefing covers Google Merchant Center AI share of voice, Lululemon conversational attributes in AI Mode, Ahrefs France CTR erosion, Cited Index saturation, Miraqo sampling, and GEO as a job function.
Paid AI discovery is leaving CPM theater. The operating question is no longer only whether you are cited. It is whether you can run a branded agent conversation and send revenue events back into the same attribution stack marketers already use for Google, Meta, and LinkedIn.
They turn an ad click into a branded AI conversation inside ChatGPT. Selected advertisers can let a user click a ChatGPT ad and start a separate, clearly labeled conversation with an AI representative of the business. OpenAI also added prompt-based ad creation and advertising integrations with HubSpot and Shopify. Sponsored agents remain in limited alpha. Paris notes that CPM and CPC inventory is a weak fit in a zero-click ChatGPT environment, and that AI discovery now includes tracking branded agent conversations inside the answer environment itself.
AI discovery is now a two-part story. It's not only about organic visibility and recommendations, but it's also about paid placement and even it's gonna be about tracking branded agent conversations and the conversation experiences inside of the answer environment itself.
By becoming OpenAI's first CRM integration for ChatGPT ads. Marketers can create campaigns in HubSpot, track them alongside other channels, attribute contacts and deals to ChatGPT spend, and send lifecycle stage and form submission conversion signals back for optimization. Paris called this the top story of the day on tape: first-party CRM revenue events flowing into ChatGPT Ads so optimization can target downstream actions, with a path toward CPA and value-based bidding (his example: roughly 3× target ROAS off HubSpot pipeline). Paid AI discovery enters the same revenue attribution infrastructure used for Google Ads, Meta, and LinkedIn.
Because leading GEO platforms are expanding into marketing OS and brand perception. Profound Aim continuously uses company data to identify problems and opportunities, propose a plan, wait for human approval, create briefs and drafts, route review, publish to the CMS, and later report what changed in share of voice and citations. Peec Brand Perception scores association and market prominence 0–100, surfaces objections with source traces, and fact-checks claims. Paris's framing: GEO is brand marketing with brand-scale budgets, not a performance SEO channel, and the competitive battle is moving beyond GEO dashboards and even GEO execution.
Merchant Center AI performance insights are generally available in Australia, Canada, India, New Zealand, and the US, reporting share of voice across AI Mode and AI Overviews by discovery, consideration, and action. Separately, Google says conversational attributes submitted by Lululemon were incorporated in 50% of relevant AI Mode product recommendations during testing, and merchants are urged to enrich feeds with contextual attributes. This is not Search Console yet, but Paris expects similar share-of-voice infrastructure to land there later.
Ahrefs' France study of 963 domains found domains with AI Overviews on more than 20% of queries saw a median −23.1% CTR (panel −5.7%; adjusted −24.8%), proving rankings and impressions can hold while traffic falls. Cited Index tracks 115 AI visibility products (109 ChatGPT, 101 Perplexity, 84 AI Overviews), signaling a fragmented market where the long tail must niche. Miraqo defaults to three prompt executions per assistant for share of voice; Paris argues 30–40 runs are needed for under 2% margin of error. GeoJobs lists 64 curated geo/AI search roles (25 dedicated, 39 hybrid), evidence that GEO is becoming a career path and that a GEO-first strategy wins both GEO and SEO.
04:30. OpenAI sponsored agents: ad click → separate labeled conversation with a brand AI rep.
07:05. HubSpot becomes OpenAI's first CRM integration for ChatGPT Ads attribution.
00:20. Profound Aim: plan → brief → CMS → SoV/citation reporting after the $180M Series D.
12:00. Lululemon conversational attributes incorporated in 50% of relevant AI Mode recommendations.
13:43. Ahrefs France: median −23.1% CTR for domains with AI Overviews on >20% of queries.
The episode closes on the same rule that runs through the sponsored-agent headline: treat paid AI discovery as revenue infrastructure and conversation design, not impression inventory.
Paris: Hi, everybody. Welcome back to another episode of The GEO Show. I'm your host, Paris Childress, and The GEO Show is brought to you by GEOforge, full self-driving for AI visibility. We're on episode 35. Let's get right to today's top stories, which again is headlined by Profound, fresh off of their one hundred and eighty million dollar raise, Series D, which they announced [00:00:30] yesterday at a one point eight billion dollar valuation.
Paris: [00:00:34] Today, they introduced Aim which works continuously across company data, identifies problems and opportunities, proposes a plan, and after approval, creates the brief and draft, routes review, publishes to the CMS, and later reports what changed. Profound now says explicitly that its ambition is to [00:01:00] become the canonical AI platform for marketing, not merely an GEO platform. [00:01:08] All right, another big move by what is now the number one player in the GEO platform space.
Paris: Profound, as we have reported before, has much greater ambitions than to be, the top GEO tool. They are thinking of something much, much bigger. So this Aim product is [00:01:30] really going from A to Z. It's ingesting data, and then it's identifying problems and opportunities, proposing the plan. [00:01:40] I guess at that point, there's a human in the loop that would approve that plan. That's the first human gate.
Paris: And then after that, it will create a brief, a draft of a piece of content. It will probably then handle all the review process that's needed, depending on the organization, publish to the CMS, [00:02:00] and then later report what has changed. [00:02:02] That's also interesting. So does that piece of content change share of voice? Does that piece of content on its own get cited?
Paris: That's also what we're trying to do as well with GEOforge, is to report on the content performance at the page level. So the competitive battle here is really now moving beyond GEO dashboards and even beyond GEO execution. [00:02:27] Profound wants to own the entire [00:02:30] marketing application layer. They wanna be the AI marketing system of record All right, moving on to the next story. Another big player, Peec, moves from visibility into brand perception and fact-checking.
Paris: Peec launched Brand Perception on September 16th. It extracts attributes associated with a brand, scores both brand association and market prominence from [00:03:00] zero to 100, surfaces recurring objections, traces those objections to sources, and checks claims against facts supplied by the brand. [00:03:11] So this is again an example of a leading GEO tool moving far beyond the visibility layer and even beyond the recommendation layer.
Paris: They're looking specifically at how models are characterizing the brand, what evidence drives [00:03:30] that characterization, and whether the characterization is accurate. So this supports a theme that I've said several times before, which is that GEO is not a performance marketing channel in the way that SEO is. [00:03:47] It is absolutely a brand marketing channel, and we're seeing it here with Peec's product, which is called Brand Perception. How does AI see my brand relative to my competitors?
Paris: Are they getting [00:04:00] our story right? Are they characterizing us right? And if not, what are the sources that this misinformation is coming from so that we can fix them and act on them? [00:04:11] So it's a very interesting move, and I do think that this is why this category is gonna be much, much bigger than SEO in-- even in the tools in the SaaS space, because it's brand marketing, and brand marketing budgets are much, much greater than performance marketing budgets.
Paris: Next [00:04:30] story. OpenAI officially turns ads into branded AI conversations [00:04:36] This story is moving very quickly because they just tested it a couple of days ago. But now, yesterday, OpenAI announced sponsored agents. They s- with selected advertisers, selected advertisers can let a user click a ChatGPT ad and start a separate, clearly labeled conversation with an AI representative of the [00:05:00] business. [00:05:00] OpenAI also added prompt-based ad creation in ChatGPT.
Paris: And a new advertising integration with-- new advertising integrations with HubSpot and Shopify. Sponsored agents remain in limited alpha rather than generally available. extremely interesting story here, and we kinda saw this coming because the old model of a [00:05:30] impressions-based ad inventory priced on CPM or even click-based priced at CPC is really not relevant in a zero-click environment like ChatGPT.
Paris: [00:05:42] So now sponsored agents are appearing, and what I thought initially was that the conversation would naturally transition into a sponsored conversation with the AI assistant making full disclosure at the right time in that conversation that it it had some, [00:06:00] sponsored products or sponsored solutions to discuss with the user. [00:06:05] But here, the user initiates it, and so the user initiates the sponsored agent conversation with a click on an ad. So that's an interesting trigger point.
Paris: And then an AI representative of the business can step in. So if I'm a business advertising on ChatGPT I should have an agent that represents my business that can [00:06:30] interact inside of ChatGPT with users after the users initiate that conversation by clicking on my ad. [00:06:37] this is fascinating stuff. I do think more and more money is gonna be rushing into ChatGPT ads, and really AI discovery is now a two-part story.
Paris: It's not only about organic visibility and recommendations, but it's also about paid placement and even it's gonna be about tracking branded agent conversations [00:07:00] and the conversation experiences inside of the answer environment itself. [00:07:05] This is a very interesting development. Next up, HubSpot closes the ChatGPT ads attribution loop. Again on ChatGPT ads. HubSpot became OpenAI's first customer relationship management integration for ChatGPT ads.
Paris: Customers can now create campaigns in HubSpot, track them alongside other channels, [00:07:30] attribute contacts and deals to ChatGPT ad spend, and send lifecycle stage and form submission conversion signals back to ChatGPT for optimization. [00:07:42] I think of all the big stories so far, I think this is our top story of the day, which is first-party data from a CRM. So first integration is HubSpot integrating with ChatGPT ads.
Paris: So that ChatGPT [00:08:00] Ads can start optimizing for downstream revenue-based events in the CRM. So this is things like, as we have seen form submissions and maybe even we're going to see values of deals or opportunities being able to flow back into ChatGPT Ads so that they can optimize for in Google Ads it's called CPA or cost per action, but here it's probably gonna [00:08:30] be... [00:08:30] It, it's still gonna be tied to the action, to the revenue action.
Paris: So paid AI discovery is moving directly into the same revenue attribution infrastructure that marketing teams already use inside of Google Ads and Meta, Meta Ads and LinkedIn ads. A fascinating move that I would like to watch very closely.
Paris: [00:08:51] I presume that HubSpot is first because they wanna go for smaller businesses that tend to use HubSpot as their CRM, [00:09:00] and clearly when I get a lead in HubSpot, I would rather just pay ChatGPT for my leads on a cost per lead type of basis, or even better, I would like to just give them a target return on ad spend and say that my sales pipeline in HubSpot is gonna be fed back into ChatGPT Ads, and basically anytime you can deliver probability weighted revenue in my sales pipeline I basically want a, a target return on ad spend on that amount of, [00:09:30] let's say, three X.
Paris: [00:09:31] The same way that you can do value-based bidding inside of Google Ads. It's a very, very savvy move by OpenAI. Combining that with the agent conversations, and we've got something extremely novel in the digital ad space. Next story is about Google. Google makes AI share of voice native in Merchant Center. [00:09:55] Google made AI performance insights generally available in Merchant Center [00:10:00] across Australia, Canada, India, New Zealand and the US.
Paris: The feature compares a retailer's share of voice with other brands across surfaces including AI Mode and AI Overviews [00:10:12] I took a look at this feature, and for the first time, Google is using and reporting on the term share of voice, and they're actually reporting on share of voice at the top of the funnel, I think it's called discovery. At the middle of the funnel, which is [00:10:30] called, I believe consideration, and then at the bottom of the funnel for action.
Paris: [00:10:35] So apparently, what they're able to do is classify the prompts or the queries coming in through Google AI Overviews and Google AI Mode. So they're classifying these prompts as basically kinda early stage discovery, middle stage consideration, and bottom of funnel high intent action. And [00:11:00] then they're looking at the brand's citation share, and I wonder how they calculate share of voice. [00:11:05] I don't think it's clearly defined. It could be mentions and citations, it could be citations only.
Paris: But they are reporting on share of voice using the term share of voice for each of those stages of the funnel inside of Merchant Center. So to be clear, this is not Google Search Console yet. It started in Merchant Center for e-commerce shops. [00:11:29] But I've gotta [00:11:30] believe that this type of reporting infrastructure is next gonna be coming to Google Search Console.
Paris: And then, of course, your share of voice reporting on any other third-party platform is really going up head-to-head with Google's own share of voice reporting inside Merchant Center and hopefully in the future, Google Search Console [00:11:51] All right, next story also about Google. Google shows structured first-party attributes can influence AI recommendations. [00:12:00] Google says that during testing with Lululemon, conversational attributes submitted by the brand were incorporated in fifty percent of relevant product recommendations in AI Mode.
Paris: [00:12:11] Google recommends merchants supply richer contextual product attributes alongside their standard feeds. So right on the heels of the last, announcement about their changes and introduction of share of voice metrics in Merchant Center, they're [00:12:30] also accepting and even encouraging e-shops basically here to submit conversational attributes and richer contextual product attributes.
Paris: [00:12:44] So what we're seeing here is that these AI engines are increasingly supporting structured product attributes canonical claims, evidence, comparisons, and machine-readable factual content beyond just [00:13:00] unstructured, knowledge base type of material.
Paris: So that means that the feeds, I think the Google Merchant Center feeds are gonna become much richer, and they're gonna start to incorporate more of these conversational attributes that the brand can enrich, and that will be rewarded by a major lift in visibility and citations basically 'cause here it says a a fifty percent lift or the, the brands that did it were [00:13:30] incorporated in a fifty percent of relevant product recommendations in AI Mode. [00:13:36] All right, next story.
Paris: France provides unusually clean evidence of AI Overview click erosion. [00:13:43] Ahrefs studied nine hundred and sixty-three domains, comparing twenty-eight days before Google AI Overviews launched in France with the first nine days afterwards. Domains with AI Overviews on more than twenty percent of their queries [00:14:00] saw a median twenty-three point one percent decline in click-through rate. [00:14:04] The full panel median was minus five point seven percent.
Paris: Ahrefs calculates a minus twenty-four point eight percent effect after adjusting for its low-exposure control group Okay, this is becoming kind of an old story in the GEO world, but it has been proven very clearly with evidence by a reputable company, Ahrefs, [00:14:30] in France that launched AI Overviews, and they saw a definitive drop in click-through rate of roughly twenty-five percent.
Paris: [00:14:38] So this is proof yet again that you could still maintain rankings and you could still maintain impressions for search, all while your actual traffic is falling sharply. All right, next story. The monitoring layer is already saturated. [00:15:00] Cited Index currently tracks one hundred and fifteen AI visibility GEO products. [00:15:06] Of those, one hundred and nine claim ChatGPT tracking. So I wonder there are six that don't do ChatGPT tracking. I wonder who those are and why.
Paris: One hundred and one track Perplexity. That's interesting because Perplexity has less than one percent market share at this point. And eighty-four track Google AI Overviews. [00:15:28] Cited Index is an [00:15:30] independent directory, not an exhaustive industry census, but it is useful as a market structure signal. So this space is very, very crowded. It's growing faster than the SEO tool space ever did.
Paris: I don't think we're gonna wind up having two major winners the way that Ahrefs and Semrush took the lion's share of the market in SEO tools. [00:15:54] I think it's gonna be much more fragmented.
Paris: We already have seen leaders emerging [00:16:00] like Profound and Peec and AirOps, but then I think there's an extremely long tail, and the way that this, the way that this long tail of providers, over a hundred and fifteen providers, will survive, I think, is to niche down and to find specific audiences, use cases, and even industries that they can focus on and specialize in. [00:16:21] Next story. Miraqo, spelled M-I-R-A-Q-O, brings repeated run GEO [00:16:30] measurement into an agency SEO suite.
Paris: Miraqo combines conventional search engine optimization capabilities with visibility across Google AI Overviews, AI Mode, ChatGPT, Gemini, and Perplexity. Its documentation says prompts are sampled repeatedly, three executions per assistant by default. [00:16:53] So repeatedly in this case means just three, not thirty or forty, but three times. Still much better [00:17:00] than one, but not enough at all to get statistical significance. And anyway, they converted those executions into share of voice, rather than a simple yes/no presence.
Paris: It also exposes SEO and AI visibility data through a MCP connector to ChatGPT, Claude, Gemini, Claude Code, and Cursor. [00:17:20] Its alternative two listing was added on September 15th, updated on September 16th. This reinforces our [00:17:30] measurement principles that repeated sampling is a must when you want to have a low margin of error reported on your AI visibility and your share of voice. These guys at Miraqo are doing it three times and bragging about it, which is interesting.
Paris: [00:17:46] I think you really need to do it thirty to forty times to achieve a less than two percent margin of error according to our research. But it's a good start for them. All right. Our last [00:18:00] story is that GEO is becoming an identifiable job function Independent job board GeoJobs, and geo is not for GEO, Generative Engine Optimization. [00:18:13] This is little geo for geography.
Paris: GeoJobs currently lists 64 open geo/AI search roles, 25 dedicated geo roles, and 39 hybrid geo plus [00:18:30] SEO positions. 16 of those are remote, 31 are US-based, and nine are in Europe. This is a curated job board. It's not a complete labor market consensus. So geo clearly now is moving from an extra responsibility for an SEO manager toward dedicated ownership budgets and a career path. [00:18:53] This is great news.
Paris: I do think that we're gonna start to see more and more of SEO [00:19:00] job titles and SEO job listings shifting over to geo. I think the first wave of this is gonna be the slash, which is SEO/GEO/AI search, and then eventually GEO is going to overtake SEO in these job titles and these job listings. [00:19:23] And as I believe the-- really the way to accomplish and win in both GEO and SEO [00:19:30] is to adopt a GEO-first strategy.
Paris: And I think more and more people are waking up to this fact, and then this is why we're starting to see the job boards starting to transition. All right, that's all we've got for today. [00:19:44] Thanks for tuning in, and we'll see you all on the next one.
