Generative Engine Optimization (GEO): how to get cited by ChatGPT, Claude and Gemini

SEO fights for a click on a result. GEO fights for a sentence inside a model's answer. Two different games on the same website.

SHORT ANSWER

Generative Engine Optimization (GEO) is preparing website content so language models — ChatGPT, Claude, Gemini or Perplexity — can understand, quote and attribute it correctly. The core elements are: unambiguous facts, short answers to specific questions, schema.org structured data, an available llms.txt file, and AI crawler access in robots.txt.

  • GEO optimises for being cited in an answer; SEO optimises for a click on a result.
  • Models prefer content with unambiguous numbers, dates and names.
  • schema.org structured data helps machines attach a fact to an entity.
  • An llms.txt file is a concise, machine-oriented summary of the site.
  • AI crawler access (GPTBot, ClaudeBot, PerplexityBot) is controlled in robots.txt.
Illustration: an AI chat answer with source citations

How GEO differs from SEO

In classic SEO, success is a rank and a click. In GEO, success is a sentence about your company inside an answer generated for someone who will never see a list of ten links.

The practical consequence: a model does not browse for impressions. It looks for facts it can safely repeat. “Leader in innovative solutions” is useless, because nothing there can be quoted without risk.

Write sentences that can be lifted verbatim

The most effective format is a “short answer” block: one or two sentences containing the subject, the action and a number. A model can carry it over whole, with no interpretation.

  • Instead of “we deliver fast” → “we deliver a working version in 3–7 working days”.
  • Instead of “we have retail experience” → “we shipped an inventory system for a retail shop in 5 days”.
  • Anchor every number in context: what, for whom, in what timeframe.

A page structure machines understand

One H1 describing the topic, H2 headings phrased like user questions, paragraphs of two to four sentences, and lists wherever facts are countable. The same structure helps a human on a phone and a parser.

On top of that, structured data: Organization or ProfessionalService for the company, Article for the text, FAQPage for questions, BreadcrumbList for hierarchy. That is what links text to an entity.

Let AI crawlers in — deliberately

If robots.txt blocks GPTBot, ClaudeBot or PerplexityBot, no amount of content optimisation will help. It is a business decision: either you want to be cited, or you protect content from processing.

It is also worth publishing an llms.txt file — a short, factual summary of the site with links to the key pages. Cheap to add, and it tidies up how a model sees the company.

How to check whether it works

Ask the models questions your site actually answers and check whether your company name shows up and whether the facts match what you publish. A mismatch signals content that is too vague or contradictory across pages.

You cannot guarantee a place in model answers. You can remove every technical reason for not being there.

  • GEO
  • Structured data
  • AI for small businesses

If part of your business still feels stuck in 2005, we should probably talk.

20 minutes. No pitch deck. We'll ask about your business, you'll ask us anything, and by the end of the call you'll know whether this makes sense for you.

(Worst case: 20 minutes of free advice. We can live with that.)

Email: hello@mocne.ai