OpenAI's ChatGPT desktop browser now supports WebMCP, an experimental standard that lets websites expose structured actions directly to AI agents instead of forcing them to click and type through a page. Episode 10 of The GEO Show covers that shift alongside five more developments, including a 107-million-answer study showing only 2% of IT services citations come from brand-owned websites and first-party Google data proving AI visibility concentrates on a small handful of pages.
OpenAI just shipped WebMCP, an experimental standard that lets ChatGPT search, edit, and complete tasks directly on a website instead of clicking and scrolling through it like a human. Episode 10 of The GEO Show opens with that shift and what it means for technical GEO audits, then covers five more developments: a 107-million-answer study on where IT services citations actually come from, first-party Google data on how concentrated AI visibility really is, Reddit's dominance of Google's discussion and forum results, evidence that ChatGPT sometimes targets specific subreddits before retrieval, and an established advisory firm's entry into the GEO consultancy market.
WebMCP is an experimental, proposed standard, currently supported through site tools in ChatGPT's desktop browser, that lets a website expose structured actions, like search, edit, or complete a task, directly to an AI agent instead of forcing that agent to navigate by clicking and typing the way a human would. OpenAI is actively running a WebMCP Developer Challenge, and supported sites can let agents act on the page directly.
"Websites that want to be built for agent use shouldn't simulate how humans use them."
That reframes what "AI-friendly" means. It isn't about making a page cleaner for a model to read, it's about whether an agent can operate on the page at all. Agent readiness now belongs next to crawlability and indexability on every technical GEO and SEO audit.
Because AI answer engines are pulling from everywhere except the brand itself. Trendos analyzed 107 million AI answers across ChatGPT, Perplexity, Gemini, and Google AI Overviews and found that among top citation sources for IT and solution services, 51% were community and UGC sources, 47% were independent editorial references, and just 2% were brand-owned websites. Frequently cited names included G2, Gartner, Clutch, GoodFirms, SourceForge, and Slashdot. For B2B technology brands, the practical answer isn't publishing everything on your own domain, it's identifying your information gain and pitching it to the higher-authority third-party sites already winning those citations.
A 99-day Google Search Console analysis of an established information site recorded 365,994 Google Generative AI impressions across 136 URLs, and just 14 of those pages generated roughly 81% of all page-level AI impressions. Among pages with 1,000+ search impressions, the median AI search visibility ratio was 23%, though individual pages varied widely. It's the same 90/10 rule long observed in SEO showing up again in GEO: a small minority of pages will carry an outsized share of AI visibility, and the job is to find them and replicate what's working.
Yes, decisively. Ahrefs analyzed 145 million US Google results and found discussion and forum results appeared in 11.7% of them, with Reddit showing up in 83.9% of that feature globally, well ahead of Facebook and Quora in a distant second and third. Longer, informational queries were especially likely to trigger discussion results, and individual thread backlink counts showed little relationship with visibility. Reddit functions like the dominant search engine of the forum landscape itself.
No. A captured ChatGPT search explicitly targeted a specific subreddit, retrieved 48 Reddit threads among 71 results, and gave Reddit 6 of 8 citations. Four days earlier, a different query retrieved 84 Reddit pages and cited none of them. Reddit's recent aggregate citation decline doesn't mean Reddit is off the table, source selection appears highly dependent on the underlying buyer question, so brands need to track citation share within their own category rather than assume a blanket trend either way.
Frost & Sullivan launched an AI brand equity and market presence advisory service using Optivara's GEO platform, benchmarking brands against competitors, analyzing how AI represents them through the funnel, diagnosing external sources shaping answers, and adding analyst interpretation on top. It's a sign that GEO is moving upstream from simple share-of-voice dashboards into brand strategy and competitive intelligence territory, and that raw share-of-voice monitoring is becoming table stakes rather than a premium feature. Real differentiation now comes from vertical expertise, proprietary brand knowledge, implementation, and pipeline impact: asking customers how they heard about a brand and tracking that back through the sales pipeline.
The thread running through all six stories: GEO keeps expanding past "did we get cited." It now includes whether an agent can act on your site, whether your citations are concentrated where you think they are, and whether raw visibility monitoring is enough to compete on anymore.
Hi, everybody. Welcome back to another exciting episode of The GEO Show, brought to you by GEOforge, full self-driving for AI visibility. I'm your host, Paris Childress, the CEO and co-founder of GEOforge. All right, let's get right into today's top stories. OpenAI is turning websites into agent-operable interfaces with something called WebMCP.
This is coming to us from OpenAI's own help center. ChatGPT's desktop browser now supports site tools built on WebMCP, an experimental and proposed standard that lets websites expose structured actions directly to AI agents instead of forcing agents to navigate by clicking and typing. OpenAI is actively running a WebMCP Developer Challenge and says supported sites can let agents search, edit, and complete tasks directly.
Very interesting stuff here. So now the web is acquiring a new optimization surface. Not just can AI find us, but can AI use us? So we're really moving into this world now of agent readiness, and for people who are doing any sort of technical audits for SEO and GEO, I think agent readiness is now a major component of that.
It dovetails nicely with the theme we've been covering quite a lot on this show already, which is that the future websites need to be ready to allow agents to operate efficiently, and that doesn't necessarily mean having agents move around like humans on websites, meaning scrolling and highlighting and copy-pasting and clicking and typing and filling out forms the way a human would.
I do remember when, in the very early days of e-commerce, there were actually websites that simulated a grocery store experience where you would literally move the mouse and the cursor down a series of grocery aisles and pick things off of shelves, which is just absurd to think about. And that just reminds me of this actually. Websites should not be built really... websites that wanna be built for agent use shouldn't necessarily simulate how humans use them.
On to the next story. A one hundred and seven million answer study says that owned websites barely register as top citation sources for IT services. So this is coming to us from Search Engine Journal. A company called Trendos analyzed one hundred and seven million AI answers across ChatGPT, Perplexity, Gemini, and Google AI Overviews. Among the top citation sources for IT and solution services, fifty-one percent were community/UGC, forty-seven percent independent editorial reference sources, and just two percent were brand-owned websites.
Now, keep in mind that this sampling of a hundred and seven million answers was specifically for IT services. So frequently cited B2B sources here included G2, Gartner, Clutch, GoodFirms, SourceForge, and Slashdot. So that's kind of what you would expect probably for IT services, particularly G2, Clutch, and Gartner. I think GoodFirms basically scrapes and copies directly from those other sources.
So for B2B technology, GEO is increasingly an off-site authority problem, and the answer here is not necessarily publishing all of your content on your website, but maybe leveraging that information gain as a valuable resource that you can use to pitch other websites that have more authority than yours and have a better chance of winning the citations through partner publications rather than your own website.
All right. Next up, first-party Google data shows that AI visibility is extremely concentrated. What does this mean? A ninety-nine-day search console analysis of an established information site, we don't know which, recorded three hundred and sixty-five thousand nine hundred and ninety-four Google Generative AI impressions across one hundred and thirty-six URLs, and just fourteen of those pages generated roughly eighty-one percent of page-level AI impressions.
Among pages with one thousand plus search impressions, the median AI search visibility ratio was twenty-three percent, but individual pages varied dramatically. AI impressions are already included inside broader search impressions, so this is not an independent traffic channel.
All right, let's unpack this. There's a well-known concept from the SEO world, which is called the 90/10 rule, and I think this came from Ethan Smith's agency a few years back. And the basic premise says that 10% of your blog posts are going to be bringing in 90% of your traffic, or it might be 90% of the search impressions. And I do think we're seeing the same thing here through this data and through this study in the world of GEO, that only fourteen pages out of 136 URLs generated 81% of page-level AI impressions.
And keep in mind, these are now the new Google Generative AI impressions. So that means effectively that that's how many times a page shows up as a cited source in either Google AI Overviews or in Google AI Mode. But the same 90/10 rule seems to be applying here, or maybe it's 80/20. And what does that mean practically speaking and for optimization purposes?
I think you do have to accept that a small minority of the pages that you publish will have an outsized impact on your AI visibility. And the key is to be able to identify those pages and then extrapolate on those pages and basically use that as a roadmap for replicating that success and taking that onto other pages.
Essentially doing more of what's working. And I will plug GEOforge here because in SignalForge, which is our reporting module, we do report on the statistics at the page level. So for every page that's published through ContentForge to a website, we report on the AI crawl activity for the page, its generative AI impressions as reported in Google Search Console, and also the referral traffic that it sends.
And yes, we do see a correlation between crawl activity and impressions. So it's clear that the pages that are getting the most crawled resources are also most likely to be winning the impressions. And the reason we show that is so that we can identify the cream at the top and basically replicate that success with other pieces of content.
All right, moving on to the next story. Reddit now appears in 83.9% of Google's discussions and forums results. Ahrefs analyzed 145 million US Google results, and what they found was discussions and forums appeared in eleven point seven percent of them. Reddit appeared in eighty-three point nine percent globally within that feature.
Following in distant second and third place was Facebook and Quora. So longer and informational queries were particularly likely to trigger discussion results, and individual thread backlink counts showed little relationship with visibility. Okay, fine. What does this mean? Now, we've just seen that Reddit has fallen off a cliff in terms of its overall share of citations in ChatGPT.
What we're talking about here is something different. We're talking about Google's discussion and forum results specifically. So it's not ChatGPT, and it's not just Google results in general or Google AI results, it's Google's discussions and forums results. What this means is that Reddit still absolutely dominates the forum landscape.
It's effectively like Google's share of search engines, with about eighty-four percent share globally. And that means that really Reddit is kind of the main game in town. You can devote some resources to Quora and to some other discussion threads, but Reddit is clearly the dominating player within UGC right now.
Now, the next story is similar to this, which is that ChatGPT is sometimes targeting individual subreddits before retrieval. A captured ChatGPT search, this is coming from Search Engine Journal, explicitly targeted a subreddit called slash whatnotapp, retrieved forty-eight Reddit threads among seventy-one results, and gave Reddit six of eight citations.
Four days earlier, another query retrieved eighty-four Reddit pages but cited none of them. So Reddit's recent aggregate citation decline does not mean that Reddit is off. Source selection appears to be highly dependent on the underlying buyer question. So I still think that this adds a little bit more nuance. We can't just say overall that we'll stop working on Reddit or try to have a presence in Reddit. You still really need to analyze the citations within your category to see what has happened with that Reddit citation share.
All right, now we have the last story of the day. Frost & Sullivan has formally entered the GEO advisory market. Frost & Sullivan launched an AI brand equity and market presence advisory service yesterday using Optivara's GEO platform. The offering benchmarks brands against competitors, analyzes how AI represents them throughout the funnel, diagnoses external sources shaping answers, and adds analyst subject matter expert interpretation.
So this is really an indication of GEO moving far upstream from just simply dashboards that show share of voice. This is getting into real brand strategy and competitive intelligence territory. And it also is proving that established advisory firms, even consulting firms, are now entering this category, GEO consultancy, which is giving strategic advice for what it's finding in LLM responses.
So the implication here is that raw share of voice monitoring is unlikely to remain a premium feature. In fact, we saw two episodes ago that it's now even a free feature in the newest tools. So the differentiation here, whether for agencies or consultancies, to add real value on top of a simple share of voice dashboard is to look at the vertical expertise, the proprietary knowledge of the brand, the implementation, which means the implementation of deploying content and building citations with off-site authority, and last but not least, and probably most importantly, the pipeline impact.
And that means looking at how many people that came through your website and filled out a form for that first lead generation, answered a question, "How did you hear about us?" by saying an AI chatbot, such as ChatGPT. That can also be confirmed on the first sales call or demo call with a salesperson. That is still the only way and the best way to determine true ROI in this new channel.
Ask your customers, "How did you hear about us?" Baseline those results, look at the impact to your sales pipeline, and then track it from there. All right. That's all for episode ten. Thank you all for listening in, and we'll see you all in the next one. Have a great weekend.
