GEO: How to Get Your Website Cited by AI Assistants
In a nutshell
GEO ensures your content gets cited by ChatGPT, Claude, or Perplexity. Three technical levers – server-rendered HTML, Schema.org markup, llms.txt – plus three content levers – direct answers, solid data, external brand mentions – cover the basics.
What is Generative Engine Optimization?
Generative Engine Optimization (GEO) prepares your content so AI assistants like ChatGPT, Claude, or Perplexity understand it, classify it as trustworthy, and cite it in their answers. While classic SEO aims for top positions in result lists, GEO is about appearing inside the generated answer itself – as a source, a citation, a recommended brand. As a growing share of B2B research now starts inside an LLM, GEO increasingly decides whether your company is even mentioned when the AI gives advice.
Technical foundation: become machine-readable
Three levers are technically decisive.
First: server-rendered HTML rather than purely client-side rendering – many LLM crawlers don't execute JavaScript.
Second: precise *Schema.org* markup (Organization, Article, FAQPage, BreadcrumbList) as JSON-LD. This tells the AI who is behind the content, when it was published, and which entities it references.
Third: an llms.txt file at the site root – it works like `robots.txt`, but for LLM crawlers, and curates the most important entry points into your domain.
GEO in numbers
- 60 %
-
of B2B research now starts in an LLM rather than a classic search engine
- 5×
-
more often FAQ and list sections get embedded into generated answers than prose.
- 40 words
-
in the introduction decide whether a page gets cited as a source.
- 3 Hebel
-
Schema.org, llms.txt, and server-rendering cover 80 % of the technical foundation
Write content that gets cited
LLMs prefer to cite sources that answer directly, name numbers, and signal authority. In concrete terms: every page answers its guiding question in the first two sentences. Important claims are backed by sources, data, or your own studies. Lists, definitions, and FAQ sections get picked up disproportionately often because they embed cleanly into a generated answer. Consistent mentions of your brand in trusted external contexts (trade press, industry directories, GitHub) further increase how often you appear in answers.
Frequently asked questions about GEO
Yes. Google results are still a training source for many LLMs, and a large portion of research still runs through classic search engines. SEO and GEO complement each other – most technical measures (clean HTML, Schema.org, clear structure) pay into both.
Through three signals: AI referral traffic in analytics (sessions from ChatGPT, Perplexity, Claude), brand-mention tracking in generated answers (manual sampling or tools like Profound, Otterly), and qualitative tests with your core search queries directly inside the LLM.
No. llms.txt helps with crawling and discovery, but Schema.org markup, server-rendered content, and well-structured answers are equally important. The file is one important piece, not a complete replacement for GEO work.
Yes, the file is publicly accessible at the site root. Treat it like your sitemap: it curates your best public content for AI crawlers – it isn't an internal strategy document.