Why ranking #1 on Google no longer means AI recommends you

A customer used to find you by typing a query into Google, scanning ten blue links, and clicking a few. You optimized for that: rank in the top results and you’d get a slice of the clicks. That was SEO, and it worked because the interface was a list.

The interface is changing. When someone asks ChatGPT, Claude, Gemini, or Google’s AI Overview “who’s the best [your service] in [your city],” they don’t get ten links. They get an answer — a short paragraph that names one, maybe three businesses. There is no page two. If the model doesn’t say your name, you weren’t ranked low; you were absent.

SEO and GEO optimize for different things¶

SEO optimizes your position in a list of links. Generative Engine Optimization (GEO) optimizes whether a model names you in a generated answer. They sound similar and they are not, because the machine in the middle works differently.

A search index stores and ranks pages. A language model synthesizes an answer from sources it retrieved and decided to trust. Two consequences fall out of that:

• Page-one isn’t a guarantee. A model can rank you #1 in classic search and still recommend a competitor, because at answer-time it pulled from the source whose information was clearest and most machine-readable — not necessarily the one Google ranked highest.
• Being citable beats being clever. The model favors content it can read, parse, and attribute confidently: plain server-rendered text, explicit facts (what you do, where, for whom), structured data it can map to entities. Clever copy that reads well to a human but is buried in images, PDFs, or JavaScript is, to the engine, missing.

“But I’m already on the first page”¶

That’s the trap. Ranking is necessary visibility for the link interface and increasingly irrelevant for the answer interface. I’ve audited sites that rank beautifully and never get named by a single AI engine, because the thing the engine needs — clear, structured, crawlable facts — wasn’t there. I’ve also seen unremarkable sites get cited constantly because their information was unambiguous and easy to attribute.

The uncomfortable part: you can’t see this happening. There’s no rank tracker for “did ChatGPT mention me.” You find out when a lead says “I asked the AI and it suggested someone else,” and by then you’ve lost the moment of intent.

What this means in practice¶

You don’t throw away SEO. The blue-link interface still exists and still sends traffic. But you can no longer treat ranking as a proxy for being found, because a growing share of your buyers are asking a machine that answers instead of lists.

GEO is the work of making sure that when the machine answers, it answers with your name. It’s mechanical, not magic: crawler access, server-rendered content, structured data, an llms.txt, and facts the engine can trust. In part 2 I’ll walk through the exact signals I check when I crawl a site the way an AI engine does — the same checklist the audit runs.