GEO is brand marketing and SEO is performance marketing, which is why Paris Childress argues that agencies folding AI search back into SEO are limiting it and that GEO will dwarf SEO over time. The episode also covers Google's September 2026 spam update, Perplexity's Photon retrieval system, a robots.txt audit of the S&P 500, why self-ranked listicles rarely get vendors named, MCP servers for first-party knowledge, automotive GEO, a $14 per month GEO stack, Growthner's case studies, and a Profound data incident.
GEO will dwarf SEO. That is Paris Childress's closing argument in this episode, and it lands in a week when agencies started folding AI search back into SEO service tiers. His case: SEO is performance marketing, built to rank pages and measure clicks, visits, and conversions. GEO is brand marketing, the work of teaching AI how to position a brand in its category and among its competitors. Treating it as a layer on top of SEO shrinks it.
In this solo roundup, Paris Childress, founder of Hop AI and co-founder of GEOforge, covers ten stories: Google's September 2026 spam update, Perplexity's Photon retrieval system, a Citation Works audit of S&P 500 robots.txt files, a Machine Relations analysis of cited versus named vendors in listicles, Stravito and Eventia opening first-party knowledge to agents through MCP, CarLocal and automotive GEO, a $14 per month GEO plus SEO stack from Easy Fetcher, Growthner's GEO case studies, a Profound data incident, and agencies rebundling AI search into SEO.
Because GEO is brand marketing and SEO is performance marketing, and brand is the bigger job. On September 25, a London consultant announced AI search monitoring and optimization would be included across all SEO service tiers while arguing against standalone GEO retainers. On September 26, Accelonaut Digital announced an expanded search practice that combines SEO with AI search optimization. Both are self-interested agency announcements, so Paris treats them as positioning signals rather than proof of a category-wide buyer shift.
"So GEO, I think over time we're gonna look back and see that GEO will have dwarfed SEO." Paris Childress
His view is that bundling GEO into SEO is the wrong mental structure. He expects GEO to become a larger AI operating system, and the platforms in the space to become AI marketing systems of record.
Rarely. Getting cited is not the same as getting named. In category roundup pages cited by AI answers, a Machine Relations analysis found:
The analysis is labeled an association, not a controlled experiment. Paris's takeaway is that a listicle-heavy strategy is not enough on its own. Content structure can earn the citation, but external category recognition, such as coverage on editorial sites and G2 reviews for SaaS, appears to matter much more for turning that citation into a brand mention.
Yes. Google began its September 2026 spam update on September 24 at approximately 9:15 AM Pacific. It applies globally across all languages and may take up to two weeks to roll out. Because most prompts involve real-time search retrieval, a shift in which sites rank changes which sites get pulled into the answer and which get cited. Paris recommends annotating core updates in AI visibility tracking alongside model changes.
Retrieval infrastructure is moving too. Perplexity says its in-house Photon retrieval and ranking system now serves all production traffic, with P99 retrieval ranking latency falling from roughly 800 milliseconds to 65 milliseconds. Its fast search API preset reports 160 milliseconds P50, 230 milliseconds P95, and 68% lower estimated model plus search cost across six agentic benchmarks, all by Perplexity's own measurement.
Not really. Citation Works audited robots.txt policies across the S&P 500 and could classify 302 of the 500 corporate files. It found zero blanket blocks for OpenAI search bot and Claude search bot, two for PerplexityBot, five for GPTBot, four for ClaudeBot, and two for Perplexity-User. The study separates search discovery, training, and user-directed retrieval, and cautions that robots directives do not prove actual crawler behavior. Paris's advice is to track how each AI bot actually crawls your site and whether it is coming for training or for real-time search delivery.
Not very, if they skip sampling depth. Easy Fetcher was listed on PitchWall on September 26 with a starter plan of $14 per month billed annually that bundles AI visibility, prompt and citation tracking, SEO analytics, backlinks, audits, white label dashboards, unlimited clients, and MCP access. Paris sees the low end of the market commoditizing fast, but warns that tracking AI visibility is sampling, not rank tracking. In GEOforge's case that means 30 to 40 runs per prompt to get the margin of error below plus or minus two percentage points.
Measurement reliability is also becoming a customer-facing issue at the top of the market. Profound's status history records a September 24 incident in which Perplexity and Google AI Overviews data collection were impacted, with six hours and thirty-seven minutes of degraded service, plus Google AI Mode citation instability earlier in September.
The thread through the episode: AI visibility is earned through outside recognition and measured with care. Brands that treat GEO as its own discipline, not an SEO add-on, will be positioned for where AI search is heading.
Subscribe to The GEO Show wherever you get your podcasts, and watch the full episode on YouTube: https://youtu.be/TM2WJfs_ZOU
Paris Childress: Hi, everybody. Welcome back to another episode of The GEO Show, brought to you by GEOforge, full self-driving for AI visibility. We got some great stories on tap here today, so let's get right into it.
Paris Childress: Today's top story is about Google. Google has launched a global spam update. Google began its September 2026 spam update on September 24th at approximately nine fifteen AM Pacific. It applies globally across all languages, and Google says that the rollout may take up to two weeks. So we don't yet know the fallout from this update.
Paris Childress: It appears to be a major update, a core ranking update, but these still can absolutely affect AI visibility rankings as well. GEO measurement can move because the underlying search and retrieval source pool changes. So to remind everybody, most of the prompts involve real-time search retrieval from search results.
Paris Childress: And if those results change dramatically, rankings change dramatically in some industries because of this core update, then that will definitely reflect on which sites get pooled and ultimately then which sites will get cited. So it can have definitely downstream effects on citation rate and AI visibility.
Paris Childress: So it's important just as you annotate in your AI visibility tracking tools or charts, share of voice charts, you should always track, of course, any model changes at the model layer, but you should also annotate when there is a major core update from Google, which is happening right now. Next up, Perplexity rebuilds retrieval with Photon.
Paris Childress: Perplexity says its new in-house Photon retrieval and ranking system now serves all production traffic. It reported its P99 retrieval ranking latency fell from roughly eight hundred milliseconds to sixty-five milliseconds while storing two point five times more data per document. Its new fast search API preset reports one hundred and sixty milliseconds P50 and two hundred and thirty milliseconds P95 latency and sixty-eight percent lower estimated model plus search cost across six agentic benchmarks at comparable aggregate quality.
Paris Childress: And these are all Perplexity's own benchmarks. All right, this is getting a little bit technical, but what this is proving is efficiency and speed at the retrieval layer, at the retrieval step, which we just discussed is an extremely important step when the models go out and do real-time search in order to provide an answer to a user.
Paris Childress: So source selection here can change underneath the model without any visible model version change, and in this case, it's about speed. The retrieval infrastructure itself, in this case, Perplexity's improvement in speed, is becoming a major variable in GEO measurement. So it will be interesting to see if citation rates, AI visibility, and share of voice are affected for those who are tracking Perplexity.
Paris Childress: We don't track Perplexity because it has such a tiny market share. It's not really worth it. We track ChatGPT, Google AI Overviews, and Google AI Mode only.
Paris Childress: Next up, AI blocked is too coarse to diagnose discovery. Citation Works audited robots.txt policies across the S&P 500 and could classify 302 out of the 500 corporate files without syntax uncertainty. Among that classifiable set, it found zero blanket blocks for OpenAI search bot and Claude search bot. It found two for PerplexityBot, five for GPTBot, four for ClaudeBot, and two for Perplexity-User.
Paris Childress: The study explicitly separates search discovery, training, and user-directed retrieval and cautions that robots directives do not prove actual crawler behavior. All right. This is a sign and a lesson here that a simple AI crawler bot allowed or blocked badge can really misdiagnose a technical problem. Robots.txt is a directive, but it still doesn't substitute the need to really deeply track how all of these AI bots are crawling your website and understanding which are coming to crawl for training purposes to update its own training corpus and its own knowledge versus which ones are coming for inference and to crawl for real-time search delivery.
Paris Childress: Next topic, one of my favorites and a recurring one about roundups and listicles. Getting cited is not the same as getting named. In category roundup pages cited by AI answers, established vendors already corroborated by other category sources were named in sixty-four out of seventy-eight citing answers. Five newer vendors that put themselves first in their own roundups were named in only eight out of eighty-five answers. Publishers absent from their own lists or placed last were named in only two of ninety-two citing answers.
Paris Childress: And the publisher here is Machine Relations, and this is Machine Relations analysis, and they explicitly label this as an association, but not a controlled experiment. So the important takeaway here is that simply creating a list, a listicle or a roundup page for your category and putting yourself anywhere on that list, either first or last, is not gonna guarantee you more citations.
Paris Childress: What's important here is that vendors need to be corroborated by other category sources that happen outside of that listicle. So if you're pursuing a listicle-heavy strategy and you're not doing a lot of other outreach to build this corroboration among editorial sites and pages within your category, and that could include things like G2 reviews if you're SaaS, it's not enough.
Paris Childress: The content structure could help you earn the citation, but the external category recognition appears to be much more important for turning that citation into a brand mention. So our focus here really should be on the category recognition at a broader level.
Paris Childress: Next up, first-party knowledge becomes agent callable. Stravito launched an MCP Model Context Protocol server on September 24th that lets ChatGPT, Claude, Copilot, and custom agents retrieve permission-scoped, source-cited company research.
Paris Childress: Eventia followed on September 25th with a native MCP server offering read-write access to events, attendees, sessions, speakers, check-in, and payments. It says the setup takes under five minutes. The big takeaway here is that proprietary knowledge from a brand, especially a brand that is willing to set up MCP, is evolving from something that companies publish for AI to crawl on their website into something that agents can query directly via MCP.
Paris Childress: And I think we're gonna see more and more brands making this move by basically allowing bot access to its proprietary knowledge and its proprietary data via MCP, which is a more direct path, as opposed to publishing pages, publishing articles on its website to make that available for crawlers. So I think that's a very important move to watch.
Paris Childress: Knowledge bases should eventually expose approved facts, claims, product data, research, and provenance through machine callable interfaces such as MCPs and APIs. And it's more than just using that to ground content. Next story. Automotive GEO verticalizes around shopper intent. CarLocal expanded its automotive AI answer engine offering on September twenty-fourth.
Paris Childress: It cited Cox Automotive's Q2 study showing sixty-three percent of in-market shoppers plan to use AI during their next purchase. Thirty-six percent already use AI for vehicle research versus thirty-eight percent using automotive-specific websites, while only twenty-nine percent of dealers say they are actively adjusting or in the process of adjusting for AI-powered search.
Paris Childress: The underlying survey covered fifteen hundred and two consumers and four hundred and eighty-three dealership decision-makers. So we are seeing here real verticalization of GEO, and in this case, into the automotive sector, where it seems to be a massive opportunity. And the data is clear. Increasingly, people that are shopping for cars, that are in-market shoppers, are increasingly using AI as their core research vehicle and their starting point.
Paris Childress: And the dealerships need to catch up. They need to close that gap. So where you have verticals like this with proprietary inventory, with local market data, things like availability and transaction data, then you're in an even stronger position to build vertical GEO products as opposed to just more generic visibility trackers.
Paris Childress: So I think CarLocal here is making a very savvy play, and they're going deep into the automotive vertical here with their proprietary data. Next story. More low-cost GEO providers are coming into the market. Fourteen dollars per month now buys a broad GEO plus SEO stack.
Paris Childress: Easy Fetcher was newly listed on PitchWall on September twenty-sixth. Its starter plan is fourteen dollars per month billed annually, and it bundles AI visibility, prompt and citation tracking, SEO analytics, backlinks, audits, white label dashboards, unlimited clients, and MCP access. Its monitoring product preserves weekly runs in append-only history.
Paris Childress: So we are seeing some activity here at the low end of the AI visibility measurement market, proving that this is commoditizing very, very quickly. Here we have a new entrant with a lot of feature breadth, very cheap multi-surface monitoring, which probably has a very low level of confidence, I'm guessing.
Paris Childress: Probably has a very high margin of error, and I think they are making a strategic error by blending and mashing together all the SEO analytics like backlink audits and rank tracking with GEO metrics. But that's beside the point. It's really showing that if you're willing to have a very high margin, low confidence on your reporting, you can get into this game and offer a super cheap product.
Paris Childress: I do think it's generally doing the market a disservice because the consumers in this space, I think, don't yet fully understand and appreciate the importance of having high confidence in the reported data. Tracking AI visibility is not like rank tracking in SEO. It's more like sampling, and you really need to sample a high number of times, in our case thirty to forty runs per prompt, in order to get the margin of error below plus or minus two percentage points.
Paris Childress: But it is proof here that the low end of the market is really commoditizing a lot of the basic features. Next up, we have some fresh case study data. GEO case studies shift the sales battle towards outcomes. Growthner published two September 25th GEO case studies. One cloud GPU client reportedly moved from 30 to 120 AI brand mentions and 200 to 500 cited pages.
Paris Childress: A design SaaS client moved from 150 to 259 mentions, 500 to 1,600 cited pages, and roughly 1,000 monthly ChatGPT referrals. Direct referrals, I guess that's to their website. Growthner says it used fixed buyer prompts via DataForSEO and citation data from Semrush. The clients are unnamed and no control groups are disclosed.
Paris Childress: All right, this is some rare case study data that we don't get too much in this space, and it is very interesting. There have been some great results from this company, Growthner. We see that AI brand mentions for this particular cloud GPU client grew four X from thirty to one hundred and twenty.
Paris Childress: And I presume that they kept a fixed set of prompts, so that didn't happen just because of measuring more prompts. And cited pages went from two hundred to five hundred. I do wonder if these are all cited pages from their own website or not. That isn't clear here to me. They had similar results from a design SaaS client.
Paris Childress: And so I think really this is proving here early on that, especially with the big increase in cited pages, that means that they must be publishing a high volume of content on their website, and that is leading to higher crawl rates by AI. Those crawl rates downstream turn into citations, and the higher degree of citations are probably then affecting AI brand mentions.
Paris Childress: So I think my takeaway here is that publishing a lot of content in an attempt to answer all the questions from all of the ICPs is a critical strategy in GEO, and these case studies here seem to confirm it. The next story comes from Profound. Profound discloses another multi-surface data incident.
Paris Childress: Profound's status history records a September twenty-fourth incident in which Perplexity and Google AI Overviews data collection were impacted, affecting visibility scores and citation rates. The status page records six hours and thirty-seven minutes of degraded service. Its history separately records Google AI Mode citation instability earlier in September.
Paris Childress: So here we're actually seeing the biggest player in the space, Profound, reporting on instability at the model layer. So this is proof that this can occur, and this is still relatively early technology. So measurement reliability is coming more and more into focus, and it's becoming more of a customer-facing product issue, not just an under-the-hood or infrastructure concern.
Paris Childress: So it's important also to be aware of the collector health issue of the models themselves, in this case, Perplexity and Google AI. And we should be aware that missing run detection, source completeness, outages, and platform change annotations are all major factors that can impact your AI search visibility in the short term.
Paris Childress: They all need to be tracked, monitored, and annotated. And on to the last story of the day. It's about bundling SEO into GEO, which I'm not a big fan of. AI search services are being rebundled into SEO. On September twenty-fifth, London consultant Charles Travers announced AI search monitoring and optimization would be included across all SEO service tiers while arguing against standalone GEO retainers.
Paris Childress: On September twenty-sixth, Accelonaut Digital announced an expanded search practice explicitly combining SEO, AEO, and AI search optimization. Both are self-interested agency announcements, so they should be treated as positioning signals rather than proof of a category-wide buyer shift. So parts of the services market here are already trying to absorb GEO back into search, and that could really put pressure on standalone GEO agency and platform pricing.
Paris Childress: My opinion is that by associating GEO too closely with traditional legacy SEO, you are really limiting the scope and the potential of GEO. Why is that? I think fundamentally SEO is very different. SEO is performance marketing with the goal of ranking pages so that you can track and measure definitive clicks, visits, and conversions that get generated from those rankings.
Paris Childress: And that is performance marketing. It's very measurable and usually has a pretty clear ROI. GEO and AI visibility is fundamentally about brand marketing. It is about training and teaching AI about your brand and how to position your brand properly in your category and among your competitors. So GEO, I think over time we're gonna look back and see that GEO will have dwarfed SEO.
Paris Childress: But these agencies that are now trying to bundle and merge SEO into GEO are really limiting and I think misunderstanding this critical point. And they're saying that, well, GEO is probably the same fundamentally as SEO, but it has new dynamics and we're gonna treat it as a layer on top. And they may be doing this to try to really just expand their current SEO budgets and SEO investments.
Paris Childress: And most of the buyers now are making this association between SEO and GEO. But I do think it's the wrong way to think about it. It's the wrong mental structure for GEO. I think GEO is gonna be a larger AI operating system, and I think that the platforms in this space will become AI marketing systems of record.
Paris Childress: All right. That's all we've got for today. Thanks for tuning in and see you all in the next one.
