Google introduced Business Agent for YouTube ads, letting viewers ask a conversational AI about products without leaving the video. Paris Childress explains why that blurs organic and paid GEO, plus Criteo's ChatGPT Ads data, a 44-point engine gap, vendor-owned citations, Athena HQ's Rootly case, and YouTube's 22.5% share of AI Overviews.
YouTube ads are about to talk back. Google introduced Business Agent for YouTube ads, which lets viewers ask a conversational AI about a product or brand without leaving the video. In a week when Ahrefs also found YouTube in 22.5% of tracked Google AI Overview responses, the line between organic video visibility and paid video is starting to blur.
In this solo roundup, Paris Childress, founder of Hop AI and co-founder of GEOforge, covers eight stories from the week in AI search: Criteo's first sizable ChatGPT Ads data set, a 44-point visibility gap between AI engines for the same brand, Google's conversational YouTube ads, a citation index where vendor-owned sites lead the top 100 cited domains, Google's developer docs as an agent tool, Athena HQ's Rootly case study, Ahrefs' YouTube share of AI Overviews, and free GEO audit tools.
More than 2,000 brands now advertise on ChatGPT through Criteo's integration, which has expanded to 61 markets. Criteo reports click-through rates two to three times higher than comparable formats, roughly four times higher ad spend after activating prompt smart ads, and more than 80% of ad-driven traffic coming from new customers. These are Criteo-reported figures, not an independent industry benchmark, but they are the first sizable performance data set for ChatGPT Ads.
Because it hides what each engine is doing. ZenMedia published results from three 2026 campaigns covering 900 buyer intent prompts across ChatGPT, Claude, Gemini, Perplexity, and Grok. In one campaign, the same brand appeared in 15.5% of prompts on its weakest engine and 59.5% on its strongest, a 44-point gap, while the five-engine average was 45.6. In two campaigns, 85% to 87% of the prompts that surfaced the client were non-branded.
Paris's view is that a single blended share of voice number is basically meaningless. At minimum, strategy should split into two parts:
Google's Business Agent for YouTube ads uses product feeds so viewers can ask immediate questions about products or brands alongside a video ad, starting with ecommerce. Paris calls it the most important story of the day.
"Now imagine this, that YouTube ads themselves are gonna become conversational." Paris Childress
He predicts YouTube ads will start appearing not only on YouTube but also in Google AI Mode, and possibly in AI Overviews, where YouTube videos already show up constantly. That blurs the line between organic and paid, and it is another reason brands need one holistic strategy that covers both organic GEO and paid AI placements, not just ChatGPT Ads.
The answer is more nuanced than the earned-media narrative suggests. A September 25 machine relations index found that 29 of its 100 most cited domains are vendor-owned, versus 22 editorial publications and 16 market and company databases. Across the wider data set, vendor-owned domains averaged 13.9 citations versus 10.8 for editorial domains. The index describes source composition and does not prove that owned content causes higher citation performance.
Most earlier studies have shown third-party UGC and editorial sites dominating citation share, so this cuts against the grain. Paris reads it as a win for owned media: brands still need to invest seriously in content on their own websites, alongside off-site citation building.
Google released a retrieving developer knowledge agent skill that teaches AI coding assistants to choose the right developer knowledge MCP tool, validate retrieval errors, and fall back to Google's REST API when MCP is unavailable. Google's MCP server gives agents direct search and retrieval access to official documentation.
Canonical knowledge is evolving from content published on a website into structured resources that agents can invoke directly. Paris suggests tools in the space explore machine-consumable distribution such as MCP endpoints, APIs, structured claim repositories, and retrieval-ready product data.
Athena HQ reports that Rootly's citation rate rose from 3% to 30% across tracked prompts, and its overall mention rate rose from 7.5% to 18% within months. The work included documentation, content, messaging, and third-party sources. Athena HQ does not publish a control group, a fixed prompt list, a tracked engine list, or attribution that separates its work from Rootly's other marketing. Even so, the category is still sparse on case studies, and a 10X citation rate lift raises the proof bar.
Ahrefs' September 25 UGC AI visibility snapshot covered 4.2 million tracked Google AI Overview responses. YouTube appeared in 22.5%, Reddit in 10%, Wikipedia in 7%, Facebook in 6%, and LinkedIn in just 0.6%.
"YouTube is absolutely dominating AI Overviews." Paris Childress
Paris calls video the number one strategy for winning visibility in Google AI Overviews. Content workflows should repurpose written content into video as much as possible, because a brand-new video with almost no views can still win the citation if it nails the topic of the prompt.
Yes, at the audit layer. Optimize.ai is promoting a free GEO platform that checks 50-plus AI search signals, audits up to 15 internal pages, generates six deployable files including llms.txt, robots.txt, and FAQ schema, covers 12-plus AI crawlers, and includes ongoing monitoring without a paid plan. Technical AI readiness audits and basic visibility checks are becoming commodity infrastructure, so the value is moving to what comes after measurement: analysis, execution, and strategy.
The takeaway for the week: video is now both the biggest organic citation opportunity in Google's AI surfaces and, with Business Agent, a conversational ad format. Brands that plan organic and paid video together will be ready when those ads reach AI Mode.
Subscribe to The GEO Show wherever you get your podcasts, and watch the full episode on YouTube: https://youtu.be/fDy-3OLlxYw
Paris Childress: Hey, everybody. Welcome to episode 47 of "The GEO Show," brought to you by GEOforge, full self-driving for AI visibility. We've got some great stories to cover on this Saturday, so let's get right into it. First up, ChatGPT Ads crosses 2,000 brands through Criteo. Criteo's September 25th update says more than 2,000 brands are now advertising on ChatGPT through its integration, which has expanded to 61 markets.
Paris Childress: Criteo reports click-through rates of two to three times comparable formats, roughly four times higher spend after activating prompt smart ads, and more than 80% of ad-driven traffic coming from new customers. So these are Criteo-reported performance figures.
Paris Childress: They're not an independent industry benchmark, but this is the first real sizable data set that we're getting back from ChatGPT Ads from Criteo across 61 markets. So ChatGPT Ads looks like it's off to a great start. Average click-through rates two to three times higher the comparable formats, and I guess they're comparing that mostly to banner, I guess banner ad, image ad formats, like in Google Display Network and maybe even social.
Paris Childress: Roughly four times higher ad spend after activating prompt smart ads, which I guess is a feature of ChatGPT Ads, and more than 80% of the ad-driven traffic coming from new customers, so that's really impressive. So this is a great start for ChatGPT Ads. All right, moving to the next story. One blended visibility score hit a 44-point engine gap.
Paris Childress: ZenMedia published results from three 2026 campaigns covering 900 buyer intent prompts across ChatGPT, Claude, Gemini, Perplexity, and Groq. In one campaign, the same brand appeared in 15.5% of prompts on its weakest engine and 59.5% on its strongest, a 44-point gap. While the five-engine average was 45.6, in two campaigns, 85% to 87% of prompts that surfaced the client were non-branded.
Paris Childress: That's a, an, a last, the last statement there is quite interesting about being non-branded, but the, the gist of this particular headline is a massive gap across these different engines when tracking visibility. So this is a very repetitive theme on this show, which is that a single share of voice number, a blended aggregate share of voice across all engines is basically meaningless.
Paris Childress: It's really gonna hide what each exact engine is doing, and the strategies really should be optimizing engine by engine. And if nothing else, it should be a two-part strategy, optimizing for ChatGPT versus optimizing for Google. And right now, as GEOforge, the way we think about it is that optimizing for Google AI Overviews and AI mode is mostly about generating YouTube videos because YouTube dominates the citation landscape in Google's AI properties.
Paris Childress: Whereas on ChatGPT, there are almost no YouTube videos in the citations, and there it's more of an outreach strategy, a PR strategy to build citations with reputable third-party sources that includes non-YouTube UGC properties like Reddit, for example. Okay, next story. Speaking of YouTube, and this is major, and I think this is the most interesting, important story of the day.
Paris Childress: Google makes YouTube ads conversational. Google introduced Business Agent for YouTube ads, allowing viewers to interact with conversational AI alongside video ads that use product feeds and ask immediate questions about products or brands without leaving the viewing context. I haven't seen this yet. I'm like really excited to play around with this. But now imagine this, that YouTube ads themselves are gonna become conversational. So you can watch a YouTube video, a YouTube ad video, and then you can immediately start interacting with that ad.
Paris Childress: You can start having a conversation with a, an AI business agent that allows you-- And then this is gonna start with e-commerce, with products. It allows you to ask more specific questions about that product and probably help you build more confidence toward your purchase. So we spoke about YouTube really dominating the landscape of citation real estate in Google AI Overviews.
Paris Childress: Now imagine when that also becomes YouTube ads, because I would predict here that YouTube ads will start appearing not only in YouTube, but also in Google AI mode. And maybe we'll see it in AI Overviews as well because we see so many YouTube videos there anyway. So that is really blurring the lines between organic and paid.
Paris Childress: And we've said before on this show that it's very important to now start thinking about a holistic strategy that encompasses both an organic GEO strategy plus a paid strategy. And mostly up until now, we have been talking about ChatGPT ads, but now we have YouTube ads that are becoming agentic. So this is gonna be a whole lot of fun when these business agent YouTube ads start appearing. All right, our next story really cuts against the grain of where we have been positioned here on this show for quite a while.
Paris Childress: Vendor-owned sites are the largest source class among the top one hundred cited domains. A September twenty-fifth machine relations index says that twenty-nine of its one hundred most cited domains are vendor-owned versus twenty-two editorial publications and sixteen market company databases. Across the wider classified dataset, vendor-owned domains average thirteen point nine domain run citations versus ten point eight for editorial domains.
Paris Childress: This describes the observed source composition. It does not prove that owned content causes higher citation performance. So this does cut against the grain quite a bit because the common belief and most of the studies so far have supported that, vendor-owned sites are a small portion of citation share.
Paris Childress: That citation share, for the most part, has been dominated by third-party sites, such as, UGC sites and editorial sites. This is actually saying the opposite, that vendor-owned sites are actually the largest source class among the top cited domains in this particular study. So this evidence is, and current evidence like it is really more nuanced than simply saying that AI engines overwhelmel- overwhelmingly prefer earned media as a narrative.
Paris Childress: This is actually a win for owned media publications. So I think one of the critical parts of a GEO strategy, always is how much do we focus on publishing content on our own website that is gonna be there to, train AI versus how much do we wanna-- how much time, energy, resources do we wanna spend in our off-site work building citations and seeding our brand all around the web?
Paris Childress: And this study seems to argue that you do still need to make a big investment in your owned properties, your vendor-owned website.
Paris Childress: Next story, Google packages its own documentation as an agent callable knowledge source. Google released a retrieving developer knowledge agent skill on September 25th. It teaches AI coding assistants how to choose the appropriate developer knowledge MCP tool, validate retrieval errors, and fall back to Google's REST application programming interface when MCP is unavailable.
Paris Childress: Google's MCP server gives agents direct search and retrieval access to official documentation. So here, what we're seeing is that canonical knowledge of AI is evolving from just content published on a website into more structured resources that agents can invoke directly. So tools in our space should really start to think about and explore more broadly machine consumable distribution like MCP endpoints, APIs, structured claim repositories, and retrieval-ready product data, not just documents used internally that ground content generation on a website.
Paris Childress: All right. I just realized that I had my slides out of order, but this is the slide about, Google turning its documentation into an agent tool.
Paris Childress: This next story is about Athena HQ. Athena HQ's Rootly case raises the category's proof bar. Athena HQ reports that Rootly's citation rate increased from three percent to thirty percent across tracked prompts, and its overall mention rate rose from seven and a half percent to eighteen percent within months.
Paris Childress: The work included documentation, content, messaging, and third-party sources. However, Athena HQ does not publish a control group, a fixed prompt list, tracked engine list, or attribution separating its interventions from Rootly's other marketing activity. Okay, this is a major case study coming out of one of the top tools, Athena HQ, and it is impressive.
Paris Childress: Citation rate more than doubling, within months. I don't know how many months that is, but it's still impressive. And, no, I'm-- Actually, the citation rate, went from three percent to thirty percent, ten X. The mention ro-- The mention rate rose from seven and a half percent to eighteen percent. So this space is still pretty sparse on case studies, but it's great to see this case study emerging from Athena HQ.
Paris Childress: Kudos to them. All right, next up, we're back to YouTube. YouTube appears in twenty-two point five percent of Ahrefs' tracked Google AI Overview responses. All right, confirming what we said earlier. Ahrefs' September twenty-fifth UGC AI visibility snapshot covers four point two million tracked Google AI Overviews responses collected on September twenty-fourth.
Paris Childress: YouTube appeared in twenty-two point five percent. Reddit appeared in ten percent, Wikipedia seven percent, Facebook six percent, LinkedIn just point six percent. So there you have it, folks. YouTube is absolutely dominating AI Overviews. I've said it before, even earlier on this episode, it is the number one strategy right now for building citations for winning visibility in Google AI Overviews, hands down.
Paris Childress: And all of your content production workflows really should be focused on not only publishing written content, but repurposing that content into video as much as possible because you have a very good chance of, as we've seen, even leapfrogging videos that are much greater, in terms of popularity channels, social engagement, and views.
Paris Childress: You can have a brand-new video with essentially no views at all, but if it's nailing the topic of the prompt, it has a great chance of winning a citation for that prompt in Google AI Overviews. Okay, on to the last story of t- of today. Free GEO tooling is compressing the technical audit layer to zero.
Paris Childress: Optimize.ai's promoted a free GEO/AEO platform on September twenty-fifth. It, its current product says it checks 50-plus AI search signals, audits up to 15 internal pages, generates six deployable files including llms.txt, robots.txt, JSON, LD, and FAQ schema. That it covers 12-plus AI crawlers and includes ongoing monitoring without a paid plan.
Paris Childress: This is really driving the cost of measurement for brands. The price of measuring your AI visibility is quickly going to zero, and the real value layer now is gonna be what comes next after measurement: analyzing and helping brands to actually act on those, those statistics and, the execution layer and the strategy layer.
Paris Childress: Both technical AI readiness audits and, basic AI visibility checks are becoming pure commodity infrastructure at this point in this category. All right, that's all we've got for episode 47. Thanks for tuning in and see you all in the next one.
