Generative Engine Optimization (GEO)

Generative engine optimization is the work of making a business more likely to be named, cited or recommended in the answers that AI assistants write. The assistants are ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews; the answers are written from pages the assistant fetched a moment earlier, so what a site serves to a crawler is what GEO acts on.

Where the term comes from

The name is from Aggarwal et al., “GEO: Generative Engine Optimization” (KDD 2024), which tested what happens to a source’s visibility in generated answers when its page is changed: adding statistics, quotations and citations, among other edits. The changes were measurable, and the paper is the closest thing to a controlled study of the mechanism. The numbers on this page are not from it; it is cited for the mechanism.

What a generative engine needs from a site

Three things, in order. The crawler has to reach the page: robots.txt is read per product token under the longest-match rule in RFC 9309, and OpenAI’s crawler documentation names three tokens with different jobs, so blocking GPTBot changes what may be trained on and not whether ChatGPT search can cite you. The page has to contain a passage that answers the question on its own, because an assistant lifts a span, not a page. And the business has to be identifiable: one name, one address, one phone number, repeated the same way, with schema that describes the business rather than the website.

Measured against 9,483 small-business sites in September 2026, most of the work is in the second and third: 61.5% had no question-and-answer content of any kind, and 43.1% were inconsistent about who and where they are. Only 3.2% blocked AI crawlers on purpose. The full breakdown is in what 9,483 small business websites serve to an AI crawler.

GEO and SEO

Search engine optimization gets a page ranked in a list a person scrolls. GEO gets a business named inside an answer a person reads instead of scrolling. The two overlap on the fundamentals, and Google says its AI features are grounded in the same ranking systems as search, so nothing here is written for a machine at the expense of the reader. They differ in the unit of selection: a rank is assigned to a page, a citation is assigned to a passage, and a recommendation is assigned to a business. Answer engine optimization (AEO) names the same practice; this site uses GEO.

How it is measured here

Twelve questions a business’s customers actually ask are put to one assistant, Claude with live web search, and every answer is kept verbatim. The share of those answers that name the business is its visibility score, from 0 to 100, with being named first reported beside it rather than weighted in, measured again on the same questions once the fixes are live, and every month with monitoring. The method is on the methodology page.

A common question

If a site already ranks well in search, does it still need GEO work?

Ranking and being named in an answer are decided on different units. A rank attaches to a page, a citation attaches to a passage, and a recommendation attaches to a business, so a page that sits high in a list can still hold nothing an assistant is able to lift. Three conditions have to hold, in order: a crawler must be able to reach the page, the page must carry a passage that answers the question on its own, and the business must be identifiable by one name, one address and one phone number repeated the same way.

Sources

The term and the mechanism: Aggarwal et al., “GEO: Generative Engine Optimization” (KDD 2024). Crawler rules: RFC 9309 and OpenAI’s crawler documentation. The 9,483-site figures are this site’s own, computed from its audit records; sample drawn per metro and trade, not from a random frame.