Generative Engine Optimisation (GEO)

More buyers now ask an AI tool to compare vendors or explain a category before they ever visit a website. The answer they get is short, and it names a few companies. Being one of them is a new kind of shortlist place.

This is not the same as ranking well in search. A company can sit on page one of a search result and still be absent from the AI answer to the same question, because the model weighs different signals, such as clear definitions and mentions across independent sites.

It is also early. Nobody can promise a result, and the engines change. But the habits that help are mostly good practice anyway: clear writing, consistent naming and a visible record of what you do. That keeps the cost of starting low.

1. Pick ten to twenty buyer questions. Use the real ones your customers ask, such as "best tool for X" or "how do I do Y". Write them down.
2. Test them. Ask each question in the main AI tools and record whether your brand appears, how it is described and who else is named.
3. State facts plainly. On your key pages, put the definition, what the product does, who it is for and what it costs in short, direct sentences a model can lift.
4. Name your brand consistently. Use the same company name, product name and description everywhere, so the model can tell it is one entity. See entity.
5. Earn mentions elsewhere. Get listed on independent review sites, directories and in articles. A model weighs agreement across the web, not only your own claims.
6. Keep pages current. Update prices, features and dates so the model is not quoting last year.
7. Re-test on a schedule. Run the same questions monthly and track share of voice in AI over time.

Generative engine optimisation (GEO) is the practice of making your brand show up, and be named or cited, in the answers that generative AI tools write. These include ChatGPT, Perplexity, Gemini and the AI summaries inside search. It differs from classic search optimisation, which fights for a place in a list of links.

A generative engine reads many sources and writes one answer. If your brand is not among the sources it trusts, the buyer reading that answer never sees you.

GEO and answer engine optimisation describe nearly the same work. GEO leans towards how a language model assembles an answer, AEO towards the answer a user ends up reading. Treat them as one discipline with two angles, and use whichever name your team already uses.

• Dropping search work. GEO adds to search optimisation. It does not replace it.
• Testing once. Answers vary between runs and between tools. Test each question several times and compare over months.
• Stuffing pages with keywords. Models look for clear statements, not repeated phrases.
• Relying on your own site. A brand mentioned only by itself carries less weight than one mentioned across several independent sources.
• Inconsistent naming. Three versions of the company name make it harder for a model to treat them as one entity.
• Promising results. Nobody controls what a model writes. Treat it as raising your chances, not guaranteeing a mention.