GEO: how to get cited by ChatGPT and Gemini
SEO got you into Google. GEO gets you into the answer the user actually reads. What Generative Engine Optimization is, how AI assistants pick their sources, and what you can do today to get cited.
More and more people no longer ask Google — they ask an AI assistant, and read only the answer. Whoever gets mentioned in that answer exists. Whoever doesn't, simply doesn't exist for that user. GEO — Generative Engine Optimization — is the craft of earning that mention.
The short answer to the question in the title: you get cited by answering real questions clearly and directly, structuring your content for machines (headings, schema.org, llms.txt), and building authority that is visible beyond your own site. The rest of this piece makes that concrete.
How an AI assistant finds its sources
To understand what works, you need to know how the answer is produced. Modern assistants combine three routes. First, training knowledge: what the model read during training — little you can influence after the fact. Second, search grounding: for current questions the assistant searches the live web and builds its answer from the pages it finds, with citations. Third, crawlers run by AI companies that roam the web collecting knowledge and citable sources.
Routes two and three are your playing field. And they work remarkably like a RAG system: retrieve, assess, summarize, cite. What we do in our own RAG pipeline to verify sources, an AI assistant does at web scale. The pages that are easy to understand, trust and summarize win.
SEO and GEO: same foundation, different goal
GEO does not replace SEO — it builds on it. Without an indexable, fast site with a logical structure you won't be found at all, not even by AI crawlers. The difference is the goal: SEO optimizes for a position in a list, GEO for a place in the answer. That shifts the emphasis from keywords to clarity: an assistant cites the page that answers its question most clearly, not the page with the most keyword repetitions.
What demonstrably helps
1. Answer the question directly. Put the answer in the first two paragraphs, not after six hundred words of run-up. Assistants summarize; what sits at the top makes the answer. Phrase your title as the question people actually ask — exactly like this article does.
2. Structure for machines. A clear heading hierarchy, short paragraphs with one thought each, lists where things truly are lists. A model "reads" your page as text structure; the more predictable it is, the smaller the chance of being summarized incorrectly.
3. Use schema.org. Structured data (Article, FAQPage, Organization, Product) tells machines explicitly what something ís instead of making them guess. FAQ blocks are extra valuable for GEO: question and answer are already in the shape an assistant needs.
4. Publish an llms.txt. This file in your site root — the robots.txt idea, but for language models — points AI systems to your most important content, ideally with markdown variants of your pages. It costs little and instantly makes your site more digestible for agents.
5. Be consistent and verifiable. Assistants cross-reference sources. If your name, numbers and claims are identical everywhere — on your site, in registers, in third-party articles — trust rises. One fabricated number that gets debunked works exactly the other way.
6. Build authority beyond your own site. Mentions in places models trust (trade press, communities, registers) carry real weight. The best route there: publish original data or insights others wánt to cite — not more press releases.
Measuring whether it works
GEO without measurement is hoping. Periodically ask the major assistants the questions you want to be found for, and record whether you are mentioned, with which source and in what wording. Also measure how readable your site is for agents. We built the Agentic Search Optimizer for exactly that: it runs a site through 15+ checks, and in one of our projects lifted an agent-readability score from 36 to 85+ — with a concrete roadmap instead of a vague report.
Frequently asked questions
Isn't GEO just hype? The term is new, the mechanism isn't: being understandable to the machine that serves your reader. That mechanism grows more important every quarter as more traffic flows through assistants.
Do I have to choose between SEO and GEO? No — ninety percent of the work overlaps. The ten percent difference (direct answers, FAQ schema, llms.txt) is exactly where you can still grab a head start today.
How fast will I see results? Think weeks to months, not days: crawlers need to come by, indexes need to refresh. So start before your competitor does.
Conclusion
GEO is not a trick but a discipline: write for people, structure for machines, and make every claim verifiable. Get that in order today and you'll be in tomorrow's answer — wait, and you'll be reading someone else's name in it.