AI search optimisation is the workflow of measuring how AI search engines answer your customers' questions, then changing what those engines retrieve about you until the answers change. The core engines for UK customers in 2026: ChatGPT search, Google AI Overviews and Perplexity, with Microsoft Copilot worth watching where buyers live in Microsoft tools. Here is the work, step by step, so you can compare any proposal against it.
The four steps
- Measure. Take the questions your customers actually ask, run them repeatedly across engines, score every answer: recommended, mentioned or absent, you and your competitors both. Repetition is not decoration; 19% of identical runs change their verdict, so single answers are noise.
- Diagnose. For each lost question: why? Unreadable pages, blocked crawlers, prices trapped in images, an inconsistent public record, missing corroboration, or a stronger rival's evidence. The fix differs per cause, which is why diagnosis precedes any content work.
- Fix. Ship the changes, with your approval, on your site and in your public record. Most fixes are unglamorous: answer-shaped pages, real text, visible dates, consistent details, structured data.
- Re-measure. Same questions, same method, plus untouched control queries and a same-day competitor panel, so an engine-wide swing cannot masquerade as progress. This step is the proof, and it is the one most proposals quietly omit.
Timelines: retrieval-level fixes can surface in answers within weeks of re-crawl; corroboration for competitive questions takes months. Anyone quoting days is guessing.
Searched for a different name?
This discipline is young enough that the market has not settled on one name for it. The same work is sold as: AEO agency · GEO agency · AI SEO agency · LLM SEO · AI visibility agency. Same four steps under every name; a proposal that skips the re-measure step is skipping the proof.
Common questions
Which AI search engines should a UK business care about?
The core three: ChatGPT search, Google AI Overviews, and Perplexity; those are the engines our published measurement data covers. Microsoft Copilot is worth adding where your buyers work inside Microsoft tools, and Bing's index influences both ChatGPT retrieval and Copilot, which makes Bing Webmaster Tools quietly important for a channel most businesses ignore. The engines disagree with each other on 29% of questions in our data, so optimising for one is not optimising for AI search.
How is AI search optimisation different from SEO?
SEO earns positions on a results page and is measured in rankings and clicks. AI search optimisation earns a place inside a generated answer and is measured in whether the answer names you. The substrate overlaps; the measurement, diagnosis and proof do not. Strong rankings with absent answers is a common and fixable state, and it is invisible to an SEO report.
What does AI search optimisation cost?
Published UK agency rates run £1,500 to £5,000+ per month in 2026, often as an add-on to an SEO retainer. Ours are public: a measured pilot from £500 one-time, then from £700 per month, no long contract. The pilot exists so you can see measured movement before committing to anything monthly.
Step one of the four is free: the ten-question audit is a small measure run on your own market, who each engine names, on your questions, today. It either finds something worth fixing or it does not, and we tell you either way. Pricing is public: pilot from £500, monthly from £700.
Updated July 2026