BrightEdge's new Agent Edge product detects AI agent requests and serves machine-optimized content at the CDN level, after finding that 97% of the top 20,000 websites it analyzed weren't serving agents any optimized format at all. Episode 9 of The GEO Show covers that launch alongside six other developments, including new data showing AI visibility doesn't travel across borders, a 60,000-query study on whether listicles are dying, and proof that Claude and Claude Code behave like two different search engines entirely.
BrightEdge just found that 97% of the top 20,000 websites it analyzed aren't serving AI agents anything they can actually use, not a poor version of the page, none at all. Episode 9 of The GEO Show opens with that finding and the CDN-level product built to fix it, then covers six more developments: why AI visibility doesn't travel across borders the way most global brands assume, a creative agency now partnering directly with an AI visibility platform, whether listicles are dying in AI search, and proof that Claude and Claude Code retrieve information like two entirely different products.
Agent Edge is a CDN-level layer that detects when a request comes from an AI agent rather than a human browser and serves that agent a reduced, machine-oriented version of the page. BrightEdge's own analysis, reported through GlobeNewswire, found that 97% of the top 20,000 websites it analyzed weren't serving any optimized machine format at all. Early customer Arm cut its payload size by about 90% using Agent Edge, while Bloomfire reported a 30% lift in direct AI referrals.
"The next step is these agents taking actions on the human user's behalf."
The market is moving from asking whether the model will cite a brand to asking whether an agent can actually interact with its website, which pushes agent access, payload efficiency, documentation, APIs, and machine-readable delivery onto every technical GEO roadmap. E-commerce is likely to feel this first, as agents start showing up to shop on behalf of the humans they represent.
No, according to Profound's Summer 2026 Index, which analyzed 19.6 billion real user conversations. In 80% of industries with cross-market data, at least one European market had a number-one brand that had never ranked first in the US, and country-specific domains accounted for 46-59% of citations where a new regional leader emerged. International SaaS and cybersecurity companies should build country- and language-specific prompt sets, earned media strategies, and citation benchmarks per market, rather than assuming a translated version of US content carries the same visibility. That same logic extends to earned authority: creative agency Influencer partnered with Profound on August 26 to connect creator content directly with AI recommendation visibility, another sign that creator marketing, digital PR, and GEO are converging into one discipline built on third-party citations and mentions rather than owned content alone.
Because they don't behave like the same product. Profound's research, covering roughly 24,000 to 135,000 responses, found that with web search enabled, Claude searched the web in 93% of its responses versus just 13% for Claude Code, and the two products overlap on only 20% of mentioned brands. Nearly three-quarters of Claude Code's agent traffic went to documentation, informational, and pricing pages, compared with just 5% for Claude. The interface and the agent harness shape retrieval almost as much as the underlying model, which means developer-focused SaaS and cybersecurity products need a documentation layer precise enough for a coding agent to retrieve compatibility, pricing, API, and deployment facts directly.
No, but their dominance is narrowing. Kevin Indig's analysis of 60,000 US Google queries and 5.32 million results, published on Search Engine Land, found listicles still appear in the top 10 for about 55% of queries, even as their share of position one fell roughly 4 percentage points since January. 62% of vendor-authored listicles gained organic traffic, with a median gain of 38%, and B2B software remains especially listicle-heavy at 62.5% of SERPs. Google appears to be pruning weaker listicles rather than banning the format, so the format still works when it carries real methodology, testing, and differentiated evidence rather than a mass-produced "ten best X" page.
Not confirmed, but the signals are credible. Search Engine Journal documented multiple reports of mass-generated AI sites losing visibility during Google's August spam update and connected the pattern to Google's own research on detecting large-scale generated spam. Google's policy remains technology-neutral: AI content is allowed and not penalized, but creating many low-value pages primarily to manipulate rankings counts as scaled content abuse. The risk was never using AI to write. It's industrialized, commodity publishing with no real information value behind it; proprietary data, subject matter expert input, first-party evidence, and genuine editorial review are what keep AI-assisted content on the right side of that line.
The thread running through all seven stories: GEO keeps moving further from "did we get mentioned" and closer to infrastructure questions, whether an agent can read a page, act on it, find it consistently across markets, and trust what it finds. Citation was the first bar to clear. It's no longer the only one.
Hi, everybody. Welcome back to episode nine of The GEO Show. We're recording on August 27th. I'm Paris Childress, your host, the CEO and co-founder of GEOforge, full self-driving for AI visibility. Check it out at getgeoforge.com. We've got a big docket ahead, so let's get right into it. First story of the day is about Profound, and Profound's Summer 2026 Index analyzed 19.6 billion real user conversations.
In 80% of industries with cross-market data, at least one European market had a number-one brand that had never ranked first in the US. Country-specific domains accounted for 46% to 59% of citations where a new regional leader emerged.
So what is this telling us? Global GEO performance is really a misleading KPI, especially for companies that operate in many different regional markets. For international companies, particularly in SaaS and cybersecurity, my advice is to build country-specific prompt sets, including language-specific prompt sets, along with different earned media strategies and citation benchmarks per market, rather than simply translating your US content and your US prompts and assuming that visibility travels.
Next up, we've got a new product announcement from BrightEdge. BrightEdge has launched an agent optimization layer at the CDN level: Agent Edge, which detects AI agent requests and serves reduced, machine-oriented content. BrightEdge says that 97% of the top 20,000 websites it analyzed were not serving optimized machine formats.
An early customer called Arm reduced its payload size by about 90%, while Bloomfire reported a 30% lift in direct AI referrals as a result of using Agent Edge. This is coming to us from GlobeNewswire. I think this is fascinating, because it's all about optimizing not just for agents' retrievability of content, but for agents to take action and smooth the path for agents taking action on your website.
So the market is now moving beyond just "can the model cite us" toward "can the agent efficiently interact with us." This is all about adding agent access, payload efficiency, documentation, APIs, and machine-readable delivery to longer-term technical GEO roadmaps, and I do think that's where it's headed.
We've spoken about this in previous episodes: we're in the first wave of AI agents on the web, where, through the lens of GEO, we're trying to optimize an agent's ability to access, crawl, understand, and ingest our content. But that's just the first step. The next step is going to be these agents taking actions on the human user's behalf.
That's another whole new layer of optimization that websites need to start thinking about, particularly e-commerce websites, where agents in the future are going to be showing up to shop on behalf of the human users they represent. Moving on to the next story, which also involves Profound.
Profound and Influencer, a creative agency, launched Creator-First AEO. The creative agency called Influencer partnered with Profound on August 26th to launch Creator-First AEO, explicitly connecting creator content and conversations with AI recommendation visibility.
So GEO/AEO is very quickly moving into the sphere of influencer marketing. Creator marketing, digital PR, and GEO are effectively converging into one authority-building discipline. One of the most effective tactics for GEO is to build third-party citations and mentions.
And what better way to do that than to enlist the help of influencer creators, who can make videos and create other types of assets about your brand and showcase them to their audiences. For B2B, this isn't necessarily about lifestyle influencers, but my advice would be that if you're pursuing GEO, you should strongly consider influencer marketing and creator marketing as one of your channels for off-site citation building.
That means mapping the YouTubers, LinkedIn experts, podcasters, analysts, and other practitioners that AI engines are already referencing, and then trying to integrate those into your earned authority campaigns. Next up in the docket is research from Kevin Indig and Search Engine Land: a 60,000-query study saying that listicles are down, but nowhere near dead.
Kevin Indig analyzed 60,000 US Google queries and 5.32 million results, and found that listicles still appear in the top 10 for about 55% of queries. Their share at position one fell roughly four percentage points between January and August, but 62% of vendor-authored listicles gained estimated organic traffic, with a median gain of 38%.
B2B software was especially listicle-heavy: 62.5% of SERPs contained at least one listicle. So it does appear that Google is pruning weaker listicles, but not banning the format. This is a pretty lively debate happening right now in GEO, because listicles are really popular, especially in B2B and software, SaaS.
They do tend to dominate, and it's a very popular tactic to create your own listicle and then reach out to other listicles for inclusion. My advice here would be: don't just go out and mass-produce a bunch of "ten best X" pages for GEO. Keep the comparison and list content only where you can provide real methodology, testing, differentiated evidence, and credible third-party context.
So listicles are not dead. Long live the listicles, but keep them focused. Next up: top-of-funnel content is holding up unusually well in cybersecurity and SaaS. A Search Engine Land analysis compared August 2026 to August 2025's estimated traffic across 30 large publishers in cybersecurity and SaaS.
Every sampled cybersecurity and marketing/sales software domain improved its top-of-funnel traffic, while finance, healthcare, legal, and consumer tech publishers all declined. The study used Semrush estimates and a small hand-selected sample, so it's directional. What does this mean? We've long been thinking about the top of funnel as the most susceptible to the zero-click trend, with so much of that traffic being taken away by AI Overviews, AI Mode, and the general shift toward LLMs.
But what we're seeing is that it's more nuanced at the category level, and AI does not eliminate informational content equally. When you have complex, fast-changing B2B subjects, I'd advise that you should still create research depth that one generated answer from AI often can't satisfy in full.
So do not give up on your top-of-funnel content, especially if you're in cybersecurity and SaaS. There's still a place for it, and there's still a classic SEO opportunity to get those pages ranked and get traffic coming into them. Next up is about Claude and Claude Code: Claude and Claude Code behave like different search engines.
This is research coming again from Profound. They analyzed between roughly 24,000 and 135,000 responses. With web search enabled, regular Claude searched in 93% of responses, versus only 13% for Claude Code. The two products overlap on just 20% of mentioned brands.
Among the observed agent visits, nearly three-quarters of Claude Code's traffic went to documentation, informational, and pricing pages, versus just 5% for Claude. This really shows us that Claude and Claude Code are very different in how they retrieve information, and that the interface and the agent harness may matter almost as much as the underlying model itself.
So the advice here, particularly for developer-focused SaaS and cybersecurity products, is that GEO really needs a rich documentation layer: precise compatibility, pricing, API documentation, deployment, authentication, and architecture facts that coding agents can retrieve directly. A lot of the coding agents developers are using are also going to be the same answer engines that need that kind of rich detail.
So don't forget about that documentation layer, especially if you're in SaaS or cybersecurity. Our final story of the day: there are credible signals, but no confirmation yet, that Google's August spam update was directed at scaled AI SEO content. Search Engine Journal documented multiple reports of mass-generated AI sites losing visibility during the August spam update, and connected the pattern to Google research on detecting large-scale generated spam.
Google has not said the update specifically targeted AI-generated content. Its official policy remains technology-neutral: AI content is allowed. Creating many low-value pages primarily to manipulate rankings is scaled content abuse. On the one hand, AI content is allowed and not penalized.
But creating many low-value pages, most likely using AI, with the primary intent to manipulate rankings is, in Google's mind, scaled content abuse. So the risk isn't using AI. The risk is industrialized commodity publishing with little original value. Which brings us back to one of the major themes on this show.
Using AI to generate content isn't bad in and of itself, as long as that content contains high information value. That means information gain: proprietary data with subject matter expert input, first-party evidence, technical expertise, and genuine editorial review. If you can get those inputs into your content and still use AI to help build and generate it, then more power to you.
That's the way we're doing it with GEOforge, and I think that's the winning formula. With that, we'll wrap up episode nine. Thank you all for listening, and we'll see you on the next episode of The GEO Show.
