What Is Generative Engine Optimisation (GEO)?
Generative Engine Optimisation is the practice of structuring content so that AI assistants cite it when answering questions. As more research ends inside an AI answer rather than on a website, being among the cited sources becomes a distinct visibility channel from ranking.
How it differs from SEO
SEO competes for position in a ranked list. GEO competes to be one of three or four sources synthesised into an answer. The mechanics differ in one important way: models select passages, not pages.
That is why smaller sites regularly get cited above much larger competitors. A precise, self-contained answer on a small site is more extractable than a rambling one on an authoritative domain. This is the clearest current opportunity in search, and it will not stay open.
What earns a citation
Answer first. Put a direct, complete answer in the opening two or three sentences of each section. A passage that buries the answer under preamble does not get lifted.
Clean definitional sentences. "Generative Engine Optimisation is the practice of..." is quotable. "When it comes to GEO, there are several considerations" is not.
Self-contained passages. Each section should make sense without the ones around it, because that is how it will be read.
Structured comparison. Tables and parallel lists get extracted far more readily than the same information in prose.
Original data. A figure that exists only on your page gives a model a reason to cite you specifically rather than any of forty equivalent pages.
Entity consistency. The same business name, description and details across your site, your profiles and third-party mentions, so a model can resolve who you are confidently.
Crawler access comes first
You cannot be cited by a system that cannot read you. A large number of sites block AI crawlers without ever deciding to — inherited robots.txt rules, WAF configuration, or CDN bot protection.
The agents worth checking: GPTBot, OAI-SearchBot and ChatGPT-User (OpenAI), PerplexityBot, ClaudeBot and Claude-SearchBot (Anthropic), Google-Extended, Bingbot, and Applebot.
Test actual fetchability rather than reading robots.txt. WAF rules routinely block crawlers that robots.txt permits, and the failure is silent.
Blocking is a legitimate choice — if content is your product, protecting it may be right. But it should be a deliberate per-crawler decision, not something discovered a year later.
llms.txt
An emerging convention: a markdown file at your root giving models a curated map of your useful content. Adoption is early and support inconsistent, so treat it as cheap insurance rather than a lever. Implement it, keep it accurate, and be sceptical of anyone presenting it as a primary tactic.
Where models look besides your site
AI systems draw heavily on Reddit, industry forums, well-maintained directories and news coverage. Visibility on those surfaces feeds citation even when your own site is not the cited source. For many businesses, a genuine presence in the communities where their customers already discuss the problem is worth more than another blog post.
Measuring it
There is no reliable automated tracker yet. Responses vary by session, phrasing and user context, and tools claiming precision are largely inferring it.
Measure by structured manual testing: a fixed set of priority queries run monthly across AI Overviews, ChatGPT, Perplexity and Copilot. Log whether you appear, in what context, and which page was cited. It takes about an hour a month and produces data you can act on.
Does GEO replace SEO?
No. AI answers draw heavily on search indexes, so ranking well remains a strong input to being cited. GEO is an additional layer that changes how you structure and present content — not a replacement for the technical and authority work underneath it.