Blog topic
Generative Engine Optimization
How to make a site quotable by AI assistants: llms.txt, crawler access and answer-first structure.
This topic covers what changes when buyers ask an AI assistant instead of typing a query: how retrieval differs from ranking, the artifacts that make a page quotable, and how to tell whether any of it is working.
What this topic covers
Generative engine optimization is the practice of structuring a site so AI assistants can find, understand and quote it accurately. It does not replace search engine optimization — it extends it at a different level. SEO optimizes a page so it ranks; GEO optimizes a section so it gets extracted and cited when an assistant answers a question.
That single shift in the unit of optimization explains most of the practical differences. A page that ranks well can still be useless to an assistant if its answers only make sense in the context of the surrounding page, and a section that reads perfectly in isolation can be lifted into an answer even when the page around it is unremarkable.
The artifacts involved
- llms.txt — a plain-Markdown file at the site root that describes the site to language models. The proposal was published by Answer.AI in September 2024; the specification lives at llmstxt.org.
- Crawler access — AI crawlers are separate user agents from classic search crawlers, and each vendor documents its own. Blocking them in robots.txt removes you from those systems regardless of how good your content is.
- Answer-first structure — leading each section with the claim, then supporting it, so a section survives being read on its own.
- Entity consistency — naming the same thing the same way site-wide, so a retrieval system can tell that scattered mentions describe one business.
- Structured data — Schema.org markup that states relationships explicitly instead of leaving them to be inferred from prose.
An honest note on measurement
Assistant answers are not deterministic, personalise per user, and are not reported in any public rank-tracking tool, so nobody can promise you a citation or a position. What you can verify is your own side: whether your artifacts are present and correct, whether the crawlers you intend to allow are actually allowed, and whether your sections still read correctly when quoted alone. The pillar article below goes through the measurement question in detail.
Where to go next
The AI visibility page covers the artifacts as a capability, including what nobody can promise. The free llms.txt generator and robots.txt generator let you produce and inspect the two files for any site, in the browser, without signing up.
August 28, 2026
Can AI Assistants Actually Read Your Website? How to Check
Before worrying about how AI assistants describe your business, confirm they can reach and parse the site at all. A practical diagnostic you can run yourself in under an hour.
June 30, 2026
Generative Engine Optimization: What Actually Makes a Site Quotable by AI
GEO is not SEO with a new label. Here is what concretely changes when your buyers ask an AI assistant instead of a search engine — and the artifacts that make your site citable.
Frequently asked questions
- Does GEO replace SEO?
- No. Optimizing for generative AI search is still SEO. GEO adds a layer on top of the same fundamentals — crawlability, structured data, clear expertise signals — and shifts the unit of optimization from the whole page to the individual section an assistant can extract.
- What is llms.txt and do I need one?
- llms.txt is a plain-Markdown file at the root of a site that describes what the site is and points to its most useful pages, proposed by Answer.AI in September 2024 with the specification published at llmstxt.org. It is a proposal rather than a standard that any assistant is obliged to honour, so treat it as cheap and harmless rather than decisive: it takes minutes to publish and cannot hurt you.
- Which AI crawlers should I allow?
- That is a business decision, not a technical one, and the honest answer is that it depends on whether you want to appear in those systems. Each vendor documents its own user agents — OpenAI, Anthropic and Perplexity all publish theirs — so the practical step is to decide per crawler and state it explicitly in robots.txt rather than leave it to a default you never chose.
- Can anyone guarantee that an AI assistant will cite my site?
- No, and you should treat any such guarantee as a reason to walk away. Assistant answers vary between users and sessions and are not reported in public rank-tracking tools. What can be verified is that your artifacts are present and correct and that your content is structured to survive extraction.
AI builds it. You own it.
Building a complete, researched website costs nothing — you pay only to publish.