GEOforge · The GEO Show, Episode 4
Only 38% of AI Overview citations now rank in Google’s top 10. A year ago it was 76%.
Eight stories dropped in the GEO world this week, and the last one should change how you report AI visibility to every client or exec who still asks “but are we ranking #1?”
We cover all of it on Episode 4 of The GEO Show. Here’s the roundup, with the receipts.
Google launched an embeddable “Preferred Sources” button on August 20. Readers who select a site through the button make it easier for that site to surface in Top Stories, AI Overviews, and AI Mode. More than 600,000 unique sources have already been selected, and publishers can promote their own selection link through email or social.
This is about as close to a free win as GEO gets right now. Adding this button to your blog costs almost nothing and creates a direct, user-driven preference signal for your pages inside Google’s AI surfaces. We’d treat this as a default addition, not an experiment.
A new entrant called Lumo launched August 20 with free AI visibility software built on a bring-your-own-API-key model. It tracks mentions, competitors, citations, prompt responses, query fan-out, sentiment, and share of voice across ChatGPT, Gemini, Perplexity, Copilot, AI Mode, and AI Overviews.
Basic prompt and citation monitoring is now a commodity, full stop. The category’s existing players, us included, have to move up the stack: content creation grounded in a proprietary knowledge base, citation-building workflows, and analysis that spans SEO and GEO together. That’s exactly why we split GEOforge into SignalForge for tracking, ContentForge for grounded content, and SiteForge for citation-building action, rather than shipping a single dashboard.
Digiday reports Zoom and multiple agencies, including Trevant and Crispin, are auditing which creators and content receive LLM citations and feeding that data into creator selection. GEO now has a direct line into influencer marketing.
Expect a new evaluation criterion to sit alongside reach, conversions, and ROI: how much citation surface area a creator’s content generates. If you run influencer programs, start asking which of your creators’ content is actually getting pulled into AI answers, not just which ones drive clicks.
BrightEdge analyzed more than 300 million monthly US searches and found Facebook cited 19.5 million times in AI Overviews, Instagram roughly 877,000 times, and TikTok 78,000 times. Facebook leads Instagram by roughly 10–20x, which leads TikTok by another 10x.
The bigger signal here is that exact answer relevance is beating follower size. Long-tail posts from accounts without huge audiences are showing up in AI Overviews because they answer a specific question well. That means social content, especially video repurposed for LinkedIn, Facebook, Instagram, and TikTok, is no longer just a distribution layer. It’s now part of the retrieval layer too, so it needs to carry concrete facts, benchmarks, and answers that stand alone when an AI system pulls it as a source.
Search Engine Journal and Ahrefs analyzed 34 self-promotional listicles across five domains and tracked 9,886 AI answers. AI systems frequently used the list as a source without recommending its publisher. In one experiment involving Ahrefs’ own conference listicle, a competing event got recommended in 43% of the answers pulling from that list.
The listicle has been one of the most popular citation-building tactics in SaaS for years: publish a “best 10 XYZ tools,” put your brand first, and hope the AI cites you. What this data shows is that an AI system can use your list as a source and then recommend a competitor from inside it, without crediting you at all. Listicles are also declining as a share of total citations, which reads like AI systems increasingly discounting them as self-promotional.
Fractal ran 96 prompts 15 times each across ChatGPT, Gemini 2.5 Flash, and Claude Sonnet, generating 4,320 responses and more than 8,500 brand references. Only 11% of brands appeared across all three models. 77% appeared in only one.
We’ve seen this inside GEOforge’s own data too: mention and citation rates diverge sharply between ChatGPT and Google’s AI properties, with Google AI Mode and AI Overviews correlating fairly closely with each other but not with ChatGPT, and Claude and Perplexity adding even more variance. A single blended share-of-voice number hides all of that. Report AI visibility at the platform level, or you’re reporting an average of numbers that don’t move together.
Tollbit analyzed AI bots from 40 vendors across 3,906 publishers and found median scraping on European sites running four times higher than on North American sites, at a ratio of one human AI referral for every 179 bot visits. The European scrape-to-referral ratio worsened from 150-to-1 in Q1 to 227-to-1 in Q2 of this year.
Crawler volume is a poor GEO success metric on its own, and this data makes the case bluntly: bot traffic is growing far faster than the human traffic AI systems actually refer back to your site, and that gap is widest in Europe. If your GEO reporting leans on crawl volume as a proxy for AI referral traffic, this is the quarter to stop.
Ahrefs analyzed 1.4 million ChatGPT prompts and attributed 88.46% of citations to the general search index. Comparing 863,000 Google result pages against 4 million AI Overview citations, only about 38% of cited URLs ranked in the top 10 for their original query, down from 76% a year earlier.
Ranking on page one used to be the finish line. Now it’s closer to the entry fee.
That’s the story of the week. It no longer reliably matters whether you rank #1 on Google if your goal is maximizing AI Overview citations, because query fan-out goes well past page one. You still need to be indexed and competitively ranked, ideally somewhere in the first two or three pages, so real-time retrieval can find you at all. But being ranked #1 stopped being the end goal it used to be.
| A year ago | Now | |
|---|---|---|
| Cited URLs also ranking top 10 | ~76% | ~38% |
| What that implies | Rank well, get cited | Rank well, maybe get cited |
Keep your SEO fundamentals in place, and expand your topic coverage around the evidence questions AI systems fan out into. A few changes worth making this week:
GEOforge tracks your citation rate directly, by platform, so you can see where you’re actually showing up in AI answers instead of inferring it from a ranking report that’s only about half as predictive as it was last year. 14-day free trial, no sales call.
Start a free trial →Sources. Google (Preferred Sources button, launched August 20, 2026); Lumo (AI visibility tool launch, August 20, 2026); Digiday (creator marketing and AI citations); BrightEdge (analysis of 300M+ monthly US searches for AI Overview social citations); Search Engine Journal and Ahrefs (34 self-promotional listicles, 9,886 tracked AI answers); Fractal (96 prompts x 15 runs across ChatGPT, Gemini 2.5 Flash, and Claude Sonnet); Tollbit (AI bot scraping analysis across 40 vendors and 3,906 publishers); Ahrefs (1.4M ChatGPT prompts, 863K Google result pages, 4M AI Overview citations). Recorded for The GEO Show, Episode 4 (Hop AI / GEOforge).
Paris Childress is the CEO of Hop AI and creator of GEOforge, a platform that helps B2B brands get cited and recommended by AI assistants like ChatGPT, Perplexity, and Gemini. A former Google Country Manager and agency veteran with 20+ years in digital marketing, Paris is focused on helping brands win in the era of AI search.
