What is generative engine optimization (GEO)?

Generative engine optimization, or GEO, is the practice of increasing how often and how favorably AI engines such as ChatGPT, Perplexity, and Gemini name and cite a product or page when they answer a question. The term comes from a 2023 paper by Aggarwal and colleagues, GEO: Generative Engine Optimization, which tested which changes to a source page made a generative engine more likely to cite it. It is the successor habit to search engine optimization for a world where the engine writes the answer instead of listing links.

How GEO differs from SEO

SEO earns a position in a list. GEO earns a place in a paragraph. A search result can rank tenth and still get clicks. An AI answer names three products and the rest do not exist. That changes the levers. SEO rewards content on domains you control, dense with the phrases people search. Generative engines discount vendor self description and pull heavily from what other people wrote: Reddit threads, Hacker News comments, comparison posts, and documentation. So GEO looks less like publishing and more like participation, and its score is share of model rather than rank.

What works in generative engine optimization

The arXiv paper found that adding citations, quotations, and statistics to a source page raised its visibility in generative answers, while keyword stuffing did nothing or hurt. In practice for a software product, the moves that show results are plain. Be present, disclosed, and useful in the recommendation threads engines already cite. Fix the stale comparison thread that repeats your old pricing. Publish a page that answers a buying question in its first sentence. And measure on a schedule, because engine answers move when models refresh. Oetzi's share of model check exists to close that loop: it lists which threads each engine cited, so you know where to show up next.

What GEO is not

It is not seeding fake reviews or paying for mentions. Engines and platforms detect astroturfing and the penalty is lasting. It is also not a thirty day project with a guaranteed outcome. Nobody outside the labs knows the weighting functions, and a model refresh can move a ranking overnight. The durable version is the same as durable Reddit strategy: show up in real threads, help, and say who you are. The GEO vs AEO vs SEO page lays out where the neighboring terms fit.

Questions

Who coined the term generative engine optimization?

The term appears in the November 2023 arXiv paper "GEO: Generative Engine Optimization" by Pranjal Aggarwal, Vishvak Murahari, Tanmay Rajpurohit, Ashwin Kalyan, Karthik Narasimhan, and Ameet Deshpande. The paper introduced a benchmark and tested which page level changes increased visibility in generative engine responses.

Does GEO work on ChatGPT?

Partly. ChatGPT answers from model memory and, when it searches, from retrieved pages. Changes to cited pages affect the retrieval path within weeks. Changes to what the model recalls without searching take longer and depend on training refreshes. Perplexity and Google AI retrieve on almost every answer, so GEO effects appear there sooner.

How do I know if GEO is working?

Write a fixed set of buying questions, ask each engine on a schedule, and track the fraction of answers that name you. That number is share of model. Track the cited sources alongside it. If the citations start including threads you participated in and pages you published, the practice is working even before the naming rate moves.