GEO, or how to be visible when a customer asks an AI about your industry

Generative Engine Optimisation (GEO) is the work of making your business citable by AI assistants, not only rankable by search engines. When a customer asks ChatGPT or Perplexity to recommend a supplier in your industry, GEO determines whether your name is in the answer.

Why ranking and being cited are different problems

Ranking puts you in a list of links; being cited puts your words in an answer. A search engine can rank a page that says nothing quotable, because the reader clicks through and reads it. An assistant needs a passage it can lift, attribute and stand behind.

That difference changes what you optimise. Keyword density and link volume move rankings, while clear self-contained passages move citations. Your page can sit on page one and never get quoted, and most businesses are in exactly that position without knowing it.

What drives a citation

Passage-level answers

Assistants quote passages, not pages. The most reliable structural change you can make is to answer the question posed by each heading in the first two sentences underneath it, before any context or preamble.

Retrieval works that way. A passage that opens with the answer survives being pulled out of its page, while one that builds to the answer over four paragraphs falls apart.

Entity clarity

An assistant has to know which company you are before it will name you. That means consistent naming across your site and profiles, and structured data (Organization, Service, FAQPage) with sameAs pointing at every profile you control.

Inconsistency is what costs you: a different legal name on the invoice, a different city on LinkedIn, an old address in a directory. Each mismatch weakens the case that these records describe one company.

Crawler access

If AI crawlers cannot read your site, none of the above matters. Check your robots.txt for GPTBot, ClaudeBot, PerplexityBot and Google-Extended, and confirm you allow them. A default rule someone added years ago may still be blocking them.

An llms.txt at your root helps too. It gives assistants a short, structured map of your site, so they stop inferring one from your navigation.

Server-rendered content

Text that only exists after JavaScript runs is text most AI crawlers will not see. If you render your key claims client-side, the systems you want citing you cannot see them.

What to fix first

Work in this order, because each step is worthless without the one before it:

1. Access. Confirm AI crawlers are allowed in robots.txt.
2. Rendering. Confirm your content exists in the HTML, not just after hydration.
3. Structure. Add Organization, Service and FAQPage schema with accurate sameAs.
4. Passages. Rewrite so each heading is answered in its first two sentences.
5. Consistency. Make your name, address and description identical everywhere.

How to know whether it worked

Ask the assistants. Put twenty realistic customer questions to ChatGPT, Claude, Perplexity and Google’s AI Overviews, and record whether you are mentioned, what is said, and who is mentioned instead. Repeat monthly.

That loop runs slower than rank tracking and you cannot shortcut it. It measures what you care about; rank tracking measures a proxy.