What is GEO (generative engine optimisation)?

Generative engine optimisation shapes what a site publishes so that AI systems can quote, cite or recommend a brand when someone asks a question.

Put a request for advice, options or a summary to an assistant and it produces an answer rather than a page of links. GEO is the work of getting your brand and its information into that answer.

The service goes by several names. Answer engine optimisation, or AEO, is the most common alternative. AI SEO is another. The labels differ, but the underlying work overlaps considerably.

AEO does highlight a change in how people search. Someone can ask a question in a full sentence, add their own circumstances and expect the system to resolve the question for them. That makes it useful to publish facts, figures and expert information in a form that can be extracted without losing the surrounding meaning.

There is a large overlap with search engine optimisation. Technical accessibility, site structure, useful content, authority and independent references still matter. The difference comes in what you are trying to earn.

SEO generally aims to earn a position on a results page and the traffic that follows. GEO aims to earn a mention or citation inside a generated answer, where the reader may never visit a traditional results page.

GEO measurement asks whether a brand appears in AI answers, and where.

Three measures appear most often: how frequently a brand is mentioned across a set of prompts, how often one of its pages is cited as a source, and what share of the mentions in a category it receives compared with competitors.

There is a catch. The thing being measured changes from one run to another.

A search results page for a particular query and location gives you a relatively stable object to measure. An AI answer is generated again each time. The wording of the prompt, the context supplied by the user and the information available to the model can all affect the result.

That means a useful GEO measurement is a sample across repeated runs. One prompt, run once, tells you very little.

The engines differ too, so deciding which ones to track is part of the measurement approach. Of the twenty agencies we reviewed, eleven name ChatGPT in their published method, nine name Google AI Overviews, nine Perplexity, eight Gemini, four Claude and three Copilot.

The measurement tools vary just as much. Seven agencies name the instrument they use, with nine different instruments named between them.

Reporting varies too. Six of the twenty agencies explain what they measure, what they use to measure it and how often they report it. Four state none of those things publicly.

If you are buying GEO, agree the prompt set, measurement tool and reporting cadence before the work starts.