🎧 Listen to the episode:
Every few years a conversation resets a mental model. This was one. I sat down with Federico Fancinelli, founder of the AI-search visibility platform GeoSonar, who spent two years running a GEO research lab — and his findings reframe how AI search actually decides who gets mentioned.
The finding that flips the script
The hopeful story was that AI search would be a great equalizer for small brands. Federico’s research says the opposite — and the reason is a single word: entities.
Jeremy: “There’s nothing like research when it comes to SEO — we have plenty of anecdotes but we’re sparse on actual research. What’s your most surprising conclusion?”
Federico: “I thought it would be easier for smaller brands to rank in AI answers — a generational opportunity, like Google twenty years ago. The research showed the opposite. It’s not ranking blue links, it’s not ranking pages — it’s ranking entities. And bigger companies are entities by definition. They organically already have the citations and the backlinks.”
In classic SEO you could rank a single clever page and outrun a bigger competitor. In GEO your whole entity is measured against other entities, and incumbents start ahead. That’s exactly why the data in your analytics is only a proxy for the thing that really moves — recognition.
The three pillars — and why coherence wins
Entity strength lives across infrastructure, narrative, and authority, unified by coherence: say one thing on your site and another on LinkedIn, and the LLMs penalize the gap. The infrastructure layer is where classic technical work still pays — it’s the same logic behind schema markup for AI visibility in the Tactic Library.
Don’t take GEO advice from Google
Federico: “What incentive does Google have to tell you, the marketer, how to rank in their own system? Zero. Imagine Google as a bank manager — they would never tell you where the cameras are positioned so you can rob their vault.”
Jeremy: “If you’re turning to me to say trust Google, I can’t do it. I have utmost confidence in John Mueller and Gary Illyes’s ability to say things that are so specifically true that they are in practice worthless.”
His verdict is narrative control — whoever controls the narrative controls the behavior (and the ad revenue). And because every LLM reasons differently, optimizing for one doesn’t rank you in the others. It’s the same fragmentation my friend Matt Brooks at SEO Teric keeps flagging.
Reputation is the new ranking
Because you rank an entity, a repeated bad review stains the whole reputation — the LLMs, Federico says, hold a grudge. As Patrick Stox put it, the real fix is often to change the product, not the page. And genuinely useful, link-worthy assets still matter — that’s the whole premise of earning links and the SEObits library.
SEO lives inside GEO
My updated model: SEO isn’t overlapping GEO — it sits inside it. Federico agreed it was “spot on.” Everything you already do still counts, plus the external ecosystem. I’m mining conversations like this into the SEO Tactic Library.
Go deeper
Explore GeoSonar and its research lab, connect with Federico on LinkedIn, and hear the full episode on Unscripted SEO. Also mentioned: OpenSEO and my Project Kilby work on entity coherence.
Go deeper → Our pillar guide, Optimising for Generative Search Results, breaks down candidate selection, entity resolution, schema and E-E-A-T — the full playbook for getting cited by AI Overviews and AI Mode.