Definition
Generative engine optimization (GEO) is the practice of shaping content so AI search engines use it and cite it in the answers they write. The term comes from a 2023 research paper by researchers at Princeton and IIT Delhi, published at the KDD 2024 conference. In day-to-day work it means the same thing as AEO, LLMO and AI SEO.
Why does generative engine optimization matter?
Generative engine optimization matters because a growing share of questions get a written answer instead of a list of links, and only the sources the answer draws on get named. If your page is not one of them, you can rank well and still be absent from the answer.
It is not a separate discipline from SEO. Google's own documentation says there are no additional requirements for appearing in AI Overviews or AI Mode: the page has to be indexed and eligible to show with a snippet. Our plain guide to AI SEO covers the full method, so this page sticks to what the research found.
How does generative engine optimization work, according to the research?
Generative engine optimization works, in the original paper, by rewriting a source so the AI answer quotes more of it. The researchers measured visibility as how much of the answer drew on each source, weighted by where it appeared, and tested a set of rewriting methods against pages left untouched.
| Method tested | What the paper reports |
|---|---|
| Citing sources, adding quotations, adding statistics | The best performers: 30 to 40% more visibility on the position-weighted measure |
| Keyword stuffing | Little to no improvement, and it scored slightly below the untouched baseline |
| Citing sources, for a page ranked fifth | Visibility up 115.1%, while the top-ranked page fell 30.3% on average |
Two things follow. The methods that worked are the ones that make a page more useful to quote, and lower-ranked pages had the most to gain. Treat the percentages as directional: they come from the researchers' own test set-up, not from Google's live results.
What do people get wrong about generative engine optimization?
- Buying it as a new service. A GEO package that cannot say what it does differently from good SEO is SEO with a new label.
- Quoting the 40% as a promise. The paper's "up to 40%" is a best case in a lab set-up, measured on its own metric.
- Blocking the crawlers that fetch pages for answers. An AI engine cannot cite a page its search bot is not allowed to read.
Sources
- Aggarwal et al., GEO: Generative Engine Optimization (arXiv 2311.09735, KDD 2024)
- Google Search Central: AI features and your website
Last checked against these sources on 17 September 2026.