AI referral traffic is converting at nearly three times the rate of Google organic search, even as SaaS organic sessions fall by close to 29%. Episode 8 of The GEO Show covers that funnel-level shift alongside nine other developments, including a case study exposing the gap between AI citations and real brand recognition, a $26 million-funded search engine built only for AI agents, and new data on why brand accuracy has to become its own GEO KPI.
AI referral traffic doesn't behave like organic search traffic, and treating the two as interchangeable is why so many GEO reports look worse than the underlying business actually is. Episode 8 of The GEO Show opens with new data showing that a shrinking flow of organic sessions can coexist with a healthier pipeline, because the visitors AI sends convert at multiples of what organic search ever did. The rest of the episode covers nine more developments: a new search engine built only for AI agents, why off-site GEO increasingly looks like digital PR, and why citation counts, brand mentions, and factual accuracy all need to be tracked as separate outcomes.
First Page Sage's analysis of 3.4 billion sessions across 218 client sites found B2B SaaS organic sessions down 29% to 29.4% from their normalized baseline, while AI platforms now account for 11.8% of SaaS sessions. ChatGPT referrals converted to leads at 4.7%, compared with just 1.9% for Google organic traffic, roughly a 3x increase. Brainlabs' separate dataset, covering 54 advertisers, found the same shape: organic sessions down 10.5%, AI referrals up 163%, and AI-driven key events up 335%.
The buyer journey has effectively collapsed from multiple searches into one long, single-threaded conversation with an LLM, so by the time a visitor clicks through, they arrive already educated and closer to converting.
Keenable emerged from stealth on August 25 with a $26 million seed round led by Accel, with participation from Conviction Partners and angel investors. It was founded by Andrey Styskin, who previously led Yandex's search, AI, and cloud division, and Matthias Petri. Keenable says it has indexed more than 100 billion documents and is already serving AI labs and inference providers. The retrieval layer beneath AI assistants is becoming its own standalone infrastructure market, and Google and Bing may not remain the only gateways to the web for AI agents.
Featured's August report analyzed 22,881 Perplexity citations across 11,499 domains and found 34.5% came from domains with a Moz domain authority below 40. In a classified subset, editorial sources accounted for 64% of citations, while brand-owned content accounted for just 1.7%. The advice that follows: build outreach lists from the publications that actually appear in buyer prompt citations, not from generic domain-rating or domain-authority thresholds. Hop AI's own Site Forge module takes this same approach, pulling citation opportunities directly from the sources already appearing in tracked prompts.
PromptScout's dataset of 5,436 completed answers across five engines found median source counts per answer of 38 for ChatGPT, 15 for Perplexity, 7 for Google AI Overviews, 6 for Gemini, and just 3 for Bing Copilot. A specialist page has a realistic shot at a ChatGPT citation and a much longer shot at Google AI Overviews, given its limited citation slots, which is why a citation-building strategy built for one platform doesn't necessarily transfer to another.
Not according to a case study from Seek Labs, which reports one site reaching 451 citations across 310 pages, with citations rising 82.6% while brand mentions rose only 10%, a roughly 8.5-to-1 ratio. AI can repeatedly use a company's content while barely strengthening its brand presence.
"It's typically easier to get a citation than to get a brand mention, but getting a brand mention, meaning that AI is naming your brand in its answer, is way, way more valuable."
Two other stories this episode reinforce the same point. Search Engine Land reports cases where sophisticated brands were omitted from AI recommendations not because AI failed to discover them, but because the available evidence surfaced real issues: missing integrations, support problems, reliability concerns. And an Imperva rank audit that tested 1,257 factual claims across 182 AI-recommended businesses found 11% had at least one claim that directly contradicted available evidence, with another 38% of claims unsupported rather than proven false. The practical takeaway: citation share, brand mentions, and factual accuracy all need to be measured and reported as separate outcomes, not folded into one visibility score.
The common thread across all ten stories: GEO keeps rewarding teams that measure the right things separately, track platform-specific behavior instead of one blended score, and treat AI's account of their brand as something to verify, not just optimize for.
Hi, everybody. Welcome back to another episode of The GEO Show. I'm your host, Paris Childress from GEOforge, the full self-driving platform for AI visibility. Check it out at getgeoforge.com. So let's get right into our top stories of the day, and we've got a nice full docket here ahead of us with ten stories.
So let's start with number one. Two fresh datasets show the same pattern, which is less organic traffic but higher intent AI visitors. An agency called First Page Sage analyzed 3.4 billion sessions across 218 client sites, and they report that B2B SaaS organic sessions are down 29% to 29.4% from its normalized baseline, while AI platforms now account for 11.8% of SaaS sessions. So this is particularly relevant for the SaaS business.
ChatGPT referrals converted to leads at 4.7% versus just 1.9% for Google organic, so roughly a 3x increase in the conversion rate. Separately, Brainlabs data covering 54 advertisers found that organic sessions are down 10.5% and AI referrals are up 163%, and AI-driven key events are up 335%.
So the interpretation here really is that AI traffic is not replacing every lost SEO click, but the visitors that do arrive from AI-referred traffic are arriving much further down the buyer's journey, and therefore they have a much higher conversion rate, as we've seen with this data. And that's particularly true with SaaS businesses.
And really that makes a lot of sense to me, having worked with dozens and dozens of SaaS businesses. It's very clear now that a lot of that research has collapsed. The buyer journey has effectively collapsed, going from multiple searches a few years ago now to long single-threaded conversations with LLMs.
And by the time someone does arrive at your website, they are much, much more educated from that conversation and much more likely to be ready to convert. In fact, the intent on coming to your website is actually to convert. They're coming with navigational intent. So really, if I would advise against evaluating GEO primarily by its referral traffic that it brings to your website, especially with SaaS.
I think what you want to do is think about this as a journey or a funnel that starts with AI exposure of your brand, which then generates branded demand, which then will lead to qualified visits and then on to, let's say, a demo or an MQL lead capture, followed by real sales pipeline. I think that is the best way for a SaaS company to evaluate the full impact, the full funnel impact of their investments in AI and GEO.
Number two story of the day is a company called Keenable. Keenable has raised $26 million to build a search engine specifically for AI agents. Keenable emerged from stealth on August 25th and secured a $26 million seed round led by Accel. They are founded by former Yandex search AI executive Andrey Styskin and Matthias Petri. And it says that it has indexed more than 100 billion documents and is already serving AI labs and inference providers.
So it may seem crazy to be launching a new search engine now, but this company is launching a search engine specifically for AI agents, which I find to be very interesting because right now most of the real-time search that is happening with LLMs is primarily going out to Google and to Bing, the dominant search engines.
But I do think that AI, especially if AI is provided with search engines that are tailored specifically for AI agents, I do think that they would also reach out and try to get those results from other sources as well. So the retrieval layer beneath AI assistants is actually becoming a standalone infrastructure market. And I don't think it's fair to assume that Google and Bing are going to remain the only gateways to the web for agents.
All right, moving on to the next story. A company called Featured launched a GEO-to-digital PR workflow with data that challenges traditional authority targeting. Let's unpack this. Featured's August report has analyzed 22,881 Perplexity citations across 11,499 domains. It found that 34.5% of citations came from domains with Moz domain authority below 40.
In a classified subset, editorial sources represented 64% of citations, while brand-owned content represented only 1.7%. Featured simultaneously launched a GEO audit intended to identify the publisher's AI engine site and turn them into PR targets. The citation figures are Perplexity-specific and can't be generalized across all other engines.
So this is interesting. There's a lot of statistics here. 34% of citations coming from low domain authority sites with domain authority below 40. So this is not a game of chasing after high domain authority websites for your citations. That's something that we covered in the last episode. And also, interestingly, editorial sources at 64% with brand-owned content only at 1.7%. So that's kind of a signal for how much effort you should be putting into off-site GEO and citation building versus on-site content publishing.
So really, the off-site GEO game is increasingly looking like a citation-informed digital PR, not really conventional high domain authority link building. So the advice here really would be to build your outreach lists and your outreach targets from the publications that are actually appearing in buyer prompt citations, not from generic DR or DA thresholds.
And in fact, this is the approach that we take with Site Forge, which is the citation building module inside of GEOforge, which is that we are pulling all of the citation opportunities directly from the citation sources of the prompts that we're tracking so that the users can really go after those specific publications rather than just trying to chase after high domain rating, high DA websites in their category. So that's great to see that this research is supporting our approach.
Next story is about Webflow. Webflow can now let Codex and ChatGPT audit and modify websites directly. Webflow's August 24th integration brings its MCP-powered capabilities into Codex and ChatGPT. The built-in workflows include SEO and AI-answer audits, accessibility checks, broken link detection, CMS management, safe publishing, and development work.
Webflow, in my opinion, has been really at the forefront of AI-search optimization for at least the last year, and this new MCP-powered integration with Codex and ChatGPT is a really big deal. I have been using the Webflow connector in Claude for many months, and I have found it to be increasingly more powerful, not only for publishing directly into my CMS, but to run audits and checks. I can, for example, check my schema.org snippets across my entire blog and get that diagnosed and improved through the Claude MCP.
So it's more good news that this has now been extended to ChatGPT's ecosystem through its Codex MCP. So I think really what this means for the practitioners out there is that the audit work is increasingly becoming automated through MCP connectors. So to say that we audit your website and we can fix the metadata, we can fix the content, that's really losing value as an agency proposition. I think that the premium layer now becomes strategy, proprietary knowledge, experimentation, QA, and just knowing which changes really matter commercially. But the pure audit work is really now going to AI.
Okay, next story. Free GEO measurement is moving into the actual consumer AI interface. A company called BrowserAct launched a free GEO report on August 25th for ChatGPT, Claude, Gemini, and Perplexity. And unlike most API-based trackers, it executes questions inside a consumer web interface, and it preserves the prompt, the visible answer, the citations. It takes screenshots. It preserves language and market. BrowserAct explicitly describes each report as a snapshot rather than ongoing monitoring.
So I think this is a very interesting new measurement methodology because most of the tools, as far as I know now, practically 100% of the AI visibility tracking tools are API-based. And I'm pretty sure that the results and answers that you get through that API are different than what a real user would get when they actually go through a browser. So I think it's very interesting to see that this company is now taking the browser approach with BrowserAct. So the measurement methodology really is becoming a differentiator here, and I'd be interested to see how their results differ from the API-driven trackers.
All right, let's go to the next story. Buyers are funding GEO, but mostly still call it SEO. A company called Fractal has surveyed 343 US marketing decision-makers. 81% still describe AI search visibility internally as SEO, while marketers report allocating an average 24% of search and content budgets to AI visibility. Two-thirds have already used ChatGPT, Gemini, Perplexity, or similar tools to evaluate marketing vendors.
So the takeaway here is that the budget is moving across the aisle faster than the terminology, meaning the crossover from SEO into GEO. So really for practitioners, agencies, I would say lead with pitches that solve the business problem, which is SEO plus visibility in AI search. In my experience, I think it is still good essentially to bundle those two concepts together. You can explain GEO after that and how it is differentiated, but I think still the market wants solutions that will solve for both SEO, which is still very relevant, and for GEO.
All right, next story. New monitoring data shows AI engines are consuming radically different amounts of evidence. PromptScout's updated dataset covers 5,436 completed answers across five engines. Its panel found median source counts per answer of 38 for ChatGPT, 15 for Perplexity, 7 for AI Overviews, 6 for Gemini, and 3 for Bing Copilot. The tracked prompts skewed towards B2B software and local services, so the exact figures should be treated as panel-specific.
So that is quite interesting that we're talking about the number of citations that are listed as sources in an answer. So clearly, ChatGPT has provided the most with 38 on average. Perplexity number two at 15, AI Overviews is just 7, and similarly 6 for Gemini, and then only 3 for Bing Copilot. So that's kind of interesting. A specialist page may have a realistic shot at getting into ChatGPT for a citation, but a more of a long shot, let's say, for getting into Google AI Overviews, which has, on average, just a fraction of the number of citations.
So I think it's really important strategically to measure really which sources each priority engine uses for a client category. And I would say if you really want to win AI citations in Google AI Overviews with the limited shelf space that's there, I would really lean heavily into YouTube videos primarily. Whereas with ChatGPT, it's a much more varied source group of sources, so I think there you can do more of a classic PR outreach to other third-party websites. So clearly there is not one universal citation building strategy or GEO content strategy. You really have to look at model by model.
Okay, moving on to the next story here. A GEO case study exposes the gap between cited and being known. Seek Labs reports one site reaching 451 citations across 310 pages, with citations rising 82.6%, while brand mentions rose only 10%. The resulting citation to mention ratio was roughly 8.5 to 1. And now this is just one client case. It's not a controlled experiment, but what this is showing potentially is that an AI system can repeatedly use a company's content while barely strengthening the company's brand presence.
And we see this in our data as well. It's typically easier to get a citation than to get a brand mention, but getting a brand mention, meaning that AI is naming your brand in its answer, is way, way more valuable. So I think what we really advise here is to measure at least three outcomes separately. You want to measure the citation from a source citation, which typically means you're coming up as one of the listed citation sources in the right-hand panel sitting next to an AI's answer. You also want to track your brand mentions in the answer, meaning just how many times your brand is appearing at all.
And last, you want to see if you are recommended with positive sentiment. So that's also important as well. Sometimes it could be something negative said about your brand, which might be damaging. So content optimization can win the first one, which means you can win a citation, but it doesn't necessarily mean that you win the other two, meaning that you might not win the brand mention or the recommendation inside of the answer itself.
All right. Moving on to the next story. Some GEO losses cannot be fixed by marketing alone. Search Engine Land reports examples where sophisticated brands were omitted from AI recommendations because available evidence surfaced real issues such as missing integrations, support problems, reliability concerns, or product limitations, not because AI failed to discover the brand.
So the takeaway here is that GEO is also becoming a customer and a product intelligence system beyond just a recommendation engine. Sometimes the model's recommendation is telling you something that's uncomfortable, but it is useful. So a strong GEO engagement should diagnose why a particular client loses recommendations and route those findings to product or customer success or support or leadership when content alone can't solve the underlying problem.
So I think it's very interesting that you could be a very well-known brand, but if you do have issues that are surfacing, AI could tell that ugly truth. And it's important not to just contain that within marketing and to try to solve that with educating AI, but also to flow that very useful information to other departments and other parts of the business that can learn from that and fix it at the source.
All right, on to the last story of this episode, which is brand accuracy needs to become a GEO KPI, and this is building on the last update. An Imperva rank audit tested 1,257 factual claims across 182 businesses recommended by a search-enabled OpenAI configuration. 24 claims, or 1.9%, contradicted available evidence, but 20 businesses, or 11%, had at least one incorrect claim. Another 38% of claims were unsupported rather than proven false. So the study used one frozen OpenAI configuration and automated quality control, so it is a snapshot rather than a universal error rate.
So this is shining a spotlight on accuracy. Most of the GEO space right now is focused on discoverability, getting your brand to simply show up in an AI response. But factual accuracy, I think, is the other side of the coin. If you're showing up but AI is giving the wrong information about your brand, or it could be that users are even using your brand in their prompt, they're asking questions about you, it's extremely important that AI is going to give factually correct and accurate information back.
If it doesn't, let's say if it hallucinates something or if it just simply gets a data point wrong, then it can do damage to the brand. So I think not only auditing how often you appear and whether you appear as a mention versus as a citation, it's also important to really understand what AI says about things like your certifications, your compliance, your integrations, your deployment models, your pricing capabilities, security claims, and basically whether public evidence supports those statements.
And if that is incorrect, then I would suggest going to the sources in your website and around the web and trying to correct that information so that you can really shape the degree of accuracy that AI says about your brand. So keep that in mind. It's not only a game of visibility, but it is also a game of brand accuracy.
All right. With that, we've come to the end of this episode. Thanks again for tuning in, and we'll see you on the next one. Bye everybody.
