Perplexity is citing mass-generated best-software pages as buying evidence. Trelner tested 380 software categories across Perplexity, Sonar, and Sonar Pro, collecting 7,534 citations, 59.8% of them from domains outside the Tranco top 100,000, with GuideFlow cited ahead of Gartner. On episode 17 of The GEO Show, recorded Thursday, 3 September 2026, Paris Childress argues that SEO authority and GEO citation authority are diverging, so brands should audit which URLs ground competitor recommendations rather than relying on backlinks.
Perplexity is citing mass-generated best-software pages as buying evidence. Trelner tested 380 software categories across Perplexity, Sonar, and Sonar Pro and collected 7,534 citations. 59.8% came from domains outside the top 100,000 on Tranco Traffic. GuideFlow, a product demo vendor, was the third most cited domain, ahead of Gartner. Three apparently related sites held 215,128 generated best-category software pages, some titled "Facts and Grounding Page."
Paris Childress recorded episode 17 of The GEO Show on Thursday, 3 September 2026. Retrieval can reward sources with little conventional authority if they align with machine retrieval. SEO authority and GEO citation authority are diverging. The rest of the briefing follows that split through model-version visibility, recommendation versus citation, licensed data, and bot configuration.
A Perplexity citation is evidence the model used a URL to ground an answer. A Google ranking is occupancy in a classic results list. Those used to travel together. On Trelner's 380-category set, they did not. The majority of the 7,534 citations sat outside the Tranco top 100,000. Pages titled "Facts and Grounding Page" were built to look like answers. Retrieval rewarded them.
AI retrieval can reward sources with very little conventional authority, provided that they are highly aligned with machine retrieval patterns.
Paris treated this as one of the strongest examples yet that SEO authority and GEO citation authority are different things, not correlated, and diverging. The useful audit is which URLs usually ground competitor recommendations, not another pass over backlinks and domain authority.
No. GuideFlow was the third most cited domain in the Trelner set, ahead of Gartner. Classic domain authority would not have called that result. SEO authority is a web graph. GEO citation authority is a retrieval graph. They are pulling apart.
Brainlabs, on 54 clients across 19 sectors, showed organic sessions down 10.5% while AI platform referrals rose 163% and AI-driven key events rose 335%, at 1.5x the organic key-event rate. The dataset does not establish causation. Sessions are a poorer proxy than visibility, recommendation, qualified visits, and pipeline. SEO Pulse, on 1,500 commercial prompts, found engines picked the same top vendor in under 1.5% of queries; about 10% of answers both mentioned and cited a brand. SignalForge runs 40 times per prompt for under 2% margin of error. Do not blend share of voice across engines.
Recommendation. At Search Seoul, Yoonhee Choi, RTN's SEO director, argued GEO should move from get-cited toward make-AI-choose-us. Structure the relationships among brand, product, capabilities, user needs, and use cases so the model knows when to recommend the brand. Knowledge structuring and ontology beat publishing more AI-friendly text. Inside GEOforge, SignalForge weights a mention three times a citation in Share of Voice.
Kevin Indig analyzed 3,000 ranking listicles in a 60,000-query set. Publishers held 46.8% of the highest-ranking listicles versus 19% for brands. AI Overviews appeared on 93.4% of B2B queries. Freshness was the strongest stable association. Putting your own product first did not independently explain performance. Earn the recommendation from somebody else.
Google began rolling Gemini 3.8 Flash into AI Mode for Google AI Pro and Ultra users globally, per Search Engine Land. Visibility is partly model-version dependent. A prompt benchmark can shift without the brand changing anything. Record the model and date. Rerun controlled baselines after major model changes.
Adobe acquired India-based Rilo in a licensing/team deal, terms undisclosed, following Semrush. The six-person team moves into Adobe. Rilo shuts down. Yext expanded Scout from tracking into GEO / AI search visibility actions across listings, reviews, social, and its knowledge graph, plus Corvo AI for SMBs. Internal 147% visibility increase. One customer doubled inbound leads. Measurement is table stakes. Execution is the fight.
Dun & Bradstreet integrated its commercial graph with Perplexity Computer via MCP, making verified B2B identity and risk data directly accessible. GEO now includes licensed databases, APIs, and MCP, not only websites. Blackstone Innovation Investments led a $27 million Series A in Huskies, valuing it above $100 million (Wall Street Journal). A badly configured WAF can make a brand invisible. Cloudflare is part of the GEO stack.
00:00. Intro. Episode 17, September 3. Perplexity is citing mass-generated best-software sites as buying evidence.
00:38. Trelner: 380 categories, 7,534 citations, 59.8% from domains outside the Tranco top 100,000. GuideFlow third, ahead of Gartner. 215,128 generated pages.
01:42. Google AI Mode rolling Gemini 3.8 Flash. Model-version dependent visibility. Rerun baselines after model changes.
02:41. Search Seoul. Yoonhee Choi: from get-cited toward make-AI-choose-us. SignalForge: a mention has 3x the weight of a citation.
04:10. Adobe acquired Rilo. Licensing/team, terms undisclosed. Six-person team into Adobe. Rilo shuts down. Follows Semrush.
05:24. Brainlabs, 54 clients, 19 sectors: organic sessions -10.5%, AI referrals +163%, AI-driven key events +335%, AI referrals 1.5x organic key-event rate. No causation claimed.
06:40. SEO Pulse, 1,500 commercial prompts. Same top vendor in under 1.5% of queries. SignalForge: 40 runs per prompt, under 2% margin of error.
08:15. Yext Scout into actions plus Corvo AI. Internal 147% visibility increase. One customer doubled inbound leads.
09:39. Dun & Bradstreet into Perplexity Computer via MCP. GEO includes licensed databases, not only websites.
11:13. Blackstone-led $27 million Series A in Huskies. A badly configured WAF can make a brand invisible. Cloudflare is part of GEO.
12:34. Kevin Indig: 3,000 listicles, publishers 46.8% vs brands 19%. Earn the recommendation from somebody else.
13:56. Outro. Episode 17 wrap.
The throughline stays the split. Retrieval can reward a page Google would ignore.
Hi, everybody. Welcome to episode seventeen of The GEO Show. I'm your host, Paris Childress from GEOforge, full self-driving for AI visibility. Let's get into today's top stories for September third, starting with Perplexity. Perplexity is citing mass-generated best software sites as buying evidence. A company called Trelner tested three hundred and eighty software categories across Perplexity, Sonar, and Sonar Pro, collecting seven thousand five hundred and thirty-four citations.
Fifty-nine point eight percent of those citations came from domains outside of the top one hundred thousand sites, according to Tranco Traffic. GuideFlow, a product demo vendor, was the third most cited domain ahead of Gartner. Trelner also identified three apparently related sites containing two hundred and fifteen thousand one hundred and twenty-eight generated best category software pages, some explicitly titled "Facts and Grounding Page."
That's very interesting. So what this is saying is that retrieval, AI retrieval can reward sources with very little conventional authority, provided that they are highly aligned with machine retrieval patterns. So it's very interesting, and I think auditing which URLs usually ground competitor recommendations is now more important than just looking at backlinks and domain authority. This is potentially one of the strongest examples yet that SEO authority and GEO citation authority are different things and not correlated and diverging.
Next up, Google Search upgraded AI Mode to Gemini 3.8 Flash. This is coming from Search Engine Land. Google began rolling out Gemini 3.8 Flash into AI Mode for Google AI Pro and Ultra users globally. Google describes the model as materially improved in multi-step reasoning and agentic tasks.
So it's actually very interesting now that AI search visibility is partly model version dependent, especially when it comes to Google AI Overviews. So a prompt benchmark can shift without the brand changing anything, due to the fact that the underlying model has changed. So it's important also for people that are tracking this to record the model and date changes in your GEO measurement and rerun those controlled baselines after major model changes.
Next up, we've got a story from Seoul, which is spelled S-E-O-U-L, and that is GEO strategy is shifting from get cited toward make AI choose us. At Search Seoul today, RTN's SEO director, Yoonhee Choi, argued that GEO should move beyond mentions and citations toward designing the relationships among brand, product, capabilities, user needs, and use cases so that AI understands when to recommend a brand. She advocates knowledge structuring and ontology rather than simply publishing more AI-friendly text.
Okay, this is a meaningful evolution here. Supports what we've been talking about a lot on this show, which is that citations are really an intermediate metric, whereas the real objective is to get recommended. And in our own platform, GEOforge, our reporting module, SignalForge, gives much heavier weighting to getting mentioned as opposed to getting cited. So when we report on Share of Voice, a mention has three times the relative weight as a citation, because I think it is much more valuable to be mentioned rather than cited.
Next up, Adobe acquired AI marketing workflow startup Rilo. So quietly, Adobe has acquired India-based Rilo in a licensing/team transaction. Terms were undisclosed. Rilo built agentic workflows for competitor intelligence, content repurposing and distribution, and sales call analysis. More than ten thousand people had reportedly tried the product, and its six-person team, plus parts of its IP, will move into Adobe, while Rilo itself will shut down.
Very interesting that this follows Adobe's Semrush acquisition. So Adobe is assembling an increasingly broad AI marketing operating layer. So keep an eye on this. Acquisitions are happening in this space. Big players like Adobe are gobbling up medium-sized players like Semrush and also smaller startups like Rilo.
Next up, AI search is sending dramatically less volume but apparently stronger intent visitors. Brainlabs analyzed Google Analytics data from fifty-four clients across nineteen sectors. What this saw is that organic sessions fell ten point five percent, while AI platform referrals rose one hundred and sixty-three percent. AI-driven key events rose three hundred and thirty-five percent, with AI referrals generating key events at one point five x the organic search rate.
Brainlabs associates the traffic inflection with wider AI Overview adoption, but the dataset does not establish causation. So Brainlabs is a leading agency here. I'm sure this is really reliable data. It's showing that organic traffic is becoming a poorer proxy for search marketing value. And really the true reporting here that is high quality is this chain or this funnel of visibility, recommendation, qualified visits, and pipeline, rather than just judging SEO or GEO primarily by sessions alone.
Next story. Another serious GEO platform has entered the race, and its data reinforces cross-engine divergence. SEO Pulse launched an enterprise AI visibility platform after testing fifteen hundred commercial prompts. It says that different AI engines selected the exact same top vendor in under one point five percent of the queries. More than half of answers mentioned brands, while only about ten percent both mentioned and cited one. It differentiates partly by running consumer-style sessions rather than API-only tests and covers Western and Asian engines.
All right. Multi-engine and multi-market measurement is rapidly becoming table stakes here now. You really cannot blend your share of voice across all engines. I think you just get such different results. It's very probabilistic. It's very volatile. This is why our SignalForge insists on doing forty runs per prompt every time we wanna measure visibility.
And of course, this increases the API costs that we have internally, but this gets us to a less than two percent margin of error and a high confidence in what we're reporting. Most of the tools in our category are not doing that.
Next, we have Yext. Again, Yext is turning AI visibility monitoring into automated execution. Yext has expanded Scout from visibility tracking into brand-level GEO / AI search visibility and actions across listings, reviews, social, and its knowledge graph. It also launched Corvo AI for small businesses. Yext reports a one hundred and forty-seven percent visibility increase internally and doubled inbound leads for one early customer.
Great. So GEO platforms, again, we see are moving very rapidly through this phase of first measurement, then diagnosis, and into execution. And now as we've seen, measurement and even diagnosis are table stakes, and the real differentiation and the real battle among tools in this emerging category is around execution. So a dashboard-only GEO service has a very short shelf life. The value's gotta sit in proprietary evidence, off-site authority building, implementation, and of course, revenue impact.
Next up, Dun & Bradstreet is bypassing the web and feeding verified B2B data directly into Perplexity. D&B, Dun & Bradstreet, has integrated its commercial graph and risk analytics with Perplexity Computer using Model Context Protocol, MCP, making verified corporate identity, relationships, and risk information directly accessible to AI workflows.
Very interesting now that there are ways to train AI that bypass publishing content directly on your website. In fact, an increasingly important part of AI visibility may start to happen outside of the public web altogether through licensed databases, APIs, and MCP connectors.
And I've been seeing this pattern for a while now, and I've been thinking it really would be much more efficient if you could directly feed your brand knowledge to AI rather than have to publish it on your website and wait and hope for AI bots to come and crawl it and ingest it and try to figure out how to format it the right way.
This is a much more direct path. So I think that GEO really needs to now include authoritative structured databases where AI agents can obtain company product facts, not just website optimization. Next story. Blackstone has put twenty-seven million dollars into protecting websites from automated AI traffic.
Blackstone Innovation Investments led a twenty-seven million dollar Series A in cybersecurity startup Huskies, valuing it above one hundred million dollars. Huskies builds an AI gatekeeper designed to distinguish and block malicious automated web traffic. This is reported by none other than The Wall Street Journal.
So bot management is becoming intertwined with AI discoverability. Sites simultaneously want legitimate AI retrieval while blocking increasingly sophisticated malicious automation. We've even seen evidence in one of our clients already that AI bots that are impersonating OpenAI's user agent bots have been showing up and appearing to crawl our website, and we have been blocking them inside of Cloudflare.
So I think that AI crawler accessibility plus bot security configuration needs to be an integral part of technical GEO audits. A badly configured web application firewall can inadvertently make a brand invisible. So really, Cloudflare is becoming an essential tool in a GEO tech stack here.
All right, last story for today. New research into 3,000 listicles complicates the listicles are dead narrative. This is research that was published yesterday from Kevin Indig, who analyzed 3,000 ranking listicles within a broader 60,000 query data set. Publishers accounted for 46.8% of highest ranking listicles versus 19% for brands and vendors.
And AI Overviews appeared on 93.4% of B2B queries in the sample. Freshness showed the strongest stable association with high rankings. Crucially, putting your own product first in a listicle did not independently explain stronger performance after control. So it's interesting here that independent publishers got about 47% of the highest ranking listicles versus just 19% for the brands themselves.
So it's not always the best strategy to launch your own listicle with yourself as number one. But it's a mixed story here. 50 best X pages can still work in SEO, but simply generating self-serving comparison content is not really a defensible GEO advantage. And of course, we're seeing that freshness is a major factor.
So you have to combine freshness with real third-party proof, original data, and credible comparisons. And you also have to earn the recommendation from somebody else, ideally. So that wraps it up for today's stories. Thanks for listening, and we'll see you all in the next one.
