2026-01-12
What is GEO, and why it matters in 2026
GEO, defined plainly
Generative engine optimisation (GEO) is the practice of writing and structuring a website so that AI answer engines — ChatGPT, Perplexity, Google AI Overviews, Microsoft Copilot — can read it, trust it, and cite it directly when someone asks a question your business can answer.
Traditional search gives a user a list of ten blue links and lets them choose. Generative search engines skip that step: they read a handful of pages, then write an answer, and — if you're lucky — name your business as the source. GEO is the set of practices that make "if you're lucky" much more likely.
Why this matters now, not eventually
Three things are true at once:
- AI assistants are now a real source of buying research. People ask ChatGPT and Perplexity questions they used to type into Google — "who's a good service in city", "what's the difference between X and Y", "is business any good". If your business isn't structured to be read and cited, it's simply absent from that conversation.
- Almost no one has done this yet. Most UK small businesses haven't touched their site's AI-crawler access, written an
llms.txtfile, or restructured their content for citation. That's a genuine, current gap — not a hypothetical future one. - The mechanics are retrieve-then-generate. Most AI answer engines fetch real web pages and generate an answer from what they find, in something close to real time. If a crawler can't access or parse your page, it cannot cite your business — no amount of reputation or ad spend fixes that on its own.
What GEO actually involves
- Allowing the right crawlers. Bots like
GPTBot,ChatGPT-User,PerplexityBot,ClaudeBotandGoogle-Extendedneed to be explicitly permitted inrobots.txt— many sites block them by accident, or by an agency's over-cautious default. - Publishing an
llms.txtfile. A clean, structured summary of what your site is and does, in the emerging standard format AI crawlers increasingly look for. - Writing for citation. Clear factual statements, concrete numbers, plainly defined terms, and direct answers to real questions — the format an AI model can quote confidently, rather than vague marketing copy it has to interpret.
- Structuring FAQs properly. Question-and-answer content, backed by
FAQPagestructured data, is exactly the shape AI answer engines lift verbatim. - Keeping content fully server-rendered. If your key content only appears after JavaScript runs in a browser, many crawlers never see it at all.
The honest caveat
GEO doesn't replace SEO, and no one — including us — can promise a specific citation in a specific AI answer, for the same reason no one can promise a specific Google ranking: the models aren't controlled by anyone outside the companies that build them. What can be delivered is a site that is genuinely readable, trustworthy and citable — which is the only lever anyone actually has.