Generative Engine Optimization (GEO): How Businesses Appear in AI Search

What GEO Can and Cannot Do

When someone asks an AI tool a question, they often get a short written answer that draws on a few sources, sometimes with links, sometimes with a business named in the text. GEO is the work of trying to be one of those sources or names.

We cannot see inside these models, so nothing here is a claim about how any algorithm ranks or decides. It is what we observed across our own studies, plus what the platforms state publicly. Treat everything below as “here is what we saw and what the platforms say,” not a rule.

GEO is close to SEO but points at a different outcome. Classic SEO works toward a position in a list of links. GEO works toward being usable inside a generated answer: information stated clearly enough, on a page a system can access, that it can be quoted or attributed. You are not defending a single ranked position, though answers can still present sources in some order.

SEO, AEO, and GEO: Overlapping, Not Separate Worlds

The vocabulary is still settling, and the terms overlap:

Google itself describes AI Overviews and AI Mode as features of Google Search with supporting links, so “AI search” is not a separate place away from Google. The useful takeaway: the fundamentals that help a system quote you (clear on-page facts, clean structure, and being mentioned on trusted sources) are the same fundamentals behind good SEO. GEO is a layer on top of solid SEO, not a replacement for it.

Four Ways to Measure AI Search Visibility

“Showing up in AI” is not one thing. It helps to separate:

These are different results, and improving one does not guarantee another. When we report a study below, we say exactly which one we measured, because “cited” and “recommended” are not the same claim.

What We Found in Our Research

Instead of repeating other people's numbers, we ran our own studies. Read these as observations in specific verticals on dated runs, reported as association, not causation.

• Most page openings do not state a buyer's decision-facts. We scored the first roughly 100 words of 852 page and location records across seven industries on how many decision-facts appear. Some records represent different pages or locations on the same domain, so the sample should not be read as 852 independent businesses, and each industry used its own set of five criteria. About 84% scored 2 out of 5 or lower, with a weighted average near 1.75. This is a content audit of openings; on its own it measures what pages say, not whether AI names them.
• Prices are often not stated in plain text. In a separate review of 187 med spa sites across 10 US metros, 55% did not state a price as plain text on the pages we checked. Without a published price, a visitor has to ask for it separately.
• Stating facts was associated with being cited, in one slice. In an observational comparison using a proxy AI-search setup (not the standard ChatGPT app), a business's own domain appeared as a cited source in at least two of three passes for 25 of 68 domains in the group stating 3-4 facts (36.8%), versus 22 of 108 in the group stating 0-2 facts (20.4%), about 1.8 times. That is an association with domain citation in one slice, not proof that rewriting an opening causes more mentions or recommendations.

In a separate regression on 174 clinics, the richness of facts on a clinic's own page was associated with domain citation, while the review metrics we used were not statistically significant. That does not prove reviews have no effect; it means our data did not show one. Full method, samples, and dates are on our research pages.

How to Start Checking Your Own Visibility

You do not need our help to begin. A simple, repeatable check:

1. 01

   Pick a few real questions a buyer would ask an AI in your field.

2. 02

   Ask them, and write down the conditions: which tool, which mode, language, market, and date, plus the answer and any sources named. Results vary between runs, so note that.

3. 03

   Score separately: was your business mentioned? was your domain cited? Track those as two different things.

4. 04

   Check your opening. Read the first 100 words of a key page. Does it state the facts a buyer decides on, price or range, who it is for, who provides it, what is included, and the next step? This is our diagnostic check of the opening, not an estimate of citation likelihood. Choose the facts that matter to your audience; information lower on the page can also be useful. Our 100-word check does this by hand or with a ChatGPT prompt.

5. 05

   Check that AI can reach the page. A page a tool cannot access is harder to cite for a full summary, though a blocked page can still surface as a bare link from other sources. Our AI crawler check reads your domain's robots signals and reports which AI crawlers appear allowed or blocked. An external check reflects HTTP-level signals only, not a guarantee that a crawler reaches every page.

6. 06

   Repeat under similar conditions to see whether anything changed.

For the full set of tactics, in order, see our practical guide.

How We Approach GEO

For a public-facing website, we study the questions your buyers actually ask, then structure pages so the facts they decide on are stated clearly and early, where both a reader and a machine can find them. We build the technical foundations, accessible HTML, clean structure, and metadata, and we make important information reachable for search and AI.

We do not promise rankings, AI citations, or a fixed rise in visibility. Those depend on factors beyond any one page, including how each tool changes over time. What we can do is put your facts where they can be read, then measure what changes.

Sources

• Google describes AI Overviews and AI Mode as features of Google Search with supporting links: Google Search Central, AI features.
• OpenAI explains crawler access for ChatGPT Search, and notes that a blocked page can still surface as a link and title obtained from other sources: OpenAI Publishers and Developers FAQ.
• Our own studies, with method, samples, and dates: the 100-Word AI Readiness study (opening fact coverage) and which pages AI cited (domain citation). Price-in-plain-text figures come from our med spa pricing study of 187 sites across 10 US metros (August 2026).