Ask ChatGPT for "a good heat pump installer in Northern Virginia" and you don't get a list of ten links. You get one answer with three or four names. Generative engine optimization (GEO) is everything you do to be one of those names. In this article, we explain what GEO is, how generative AI models arrive at their answers and why, for small and midsize businesses, this isn't hype but a shift in how customers choose a provider.
A definition of GEO
Generative engine optimization is optimizing your business, your website and your presence on the web so that generative AI models name you, describe you correctly and recommend you. SEO is about positions in a list of search results. GEO is about mentions in a generated answer. The answer is the new search result, and there's only one answer.
The term first appeared in 2023 in a research paper by researchers at Princeton and Georgia Tech, who tested which properties of web content make a language model more likely to use that content. Since then, GEO has grown into its own discipline, with its own metrics (visibility, share of voice, citations) and its own tools.
How an AI model arrives at its answer
An AI model like ChatGPT doesn't answer a question from memory. For questions about businesses, products and services, it first runs one or more web searches (fan-outs, in industry jargon) and reads a handful of pages it trusts. From those pages, the model distills an answer, often with links to its sources (citations).
That means three things. One: the model only names businesses it finds in the sources it reads. Two: those sources are usually not your own website, but comparison sites, trade media, forums and industry associations. Three: the answer varies by model, by day and by how the question is phrased, because the searches and the chosen sources vary.
Why GEO differs from SEO
SEO and GEO share a foundation (good content, a technically healthy site, authority), but the mechanics differ. In Google, you can rank seventh and still get traffic. In an AI answer, you're either in it or you're not. Google rewards pages. AI models reward entities: businesses that are named consistently, recognizably and often in places the model trusts.
A practical difference: the page that ranks first on Google for "heat pump installer" is far from always the page ChatGPT cites. AI models prefer pages that answer a question directly and completely, with concrete facts, prices and comparisons. A long sales page does poorly there. An honest comparison page does well.
Why this matters for small businesses
Large companies get named anyway; they're in every source. For a business with five to twenty employees, the question is whether AI knows you at all. In practice, the average small business shows up in only a small share of relevant answers, while a handful of competitors per region take most of the mentions.
The good news: in local and specialized markets, you can close that gap quickly. One listing on the right comparison site, a robots.txt that lets AI crawlers in and a page that actually answers your customer's question can make a visible difference within a few weeks.
Where to start
- Measure where you stand: what questions do customers ask AI, and are you in the answer?
- Map the sources AI cites in your market. That's where your leverage is.
- Check that your website is readable for AI crawlers (robots.txt, structure, speed).
- Write content that answers questions instead of listing services.
- Measure again every week. AI answers change faster than Google rankings.
GEO doesn't replace SEO. It's a layer on top: you optimize not for a list, but for an answer. Whoever measures first knows first where the opportunities are.