GEO & AEO: how to get cited by ChatGPT, Perplexity and AI Overviews
July 28, 2026 · 7 min read
What GEO/AEO is and how AI engines pick their sources. An 8-step checklist: llms.txt, citable passages, FAQ, entities, crawler access — and how to measure results.
More and more search sessions end without a click: the answer is assembled by AI — ChatGPT, Perplexity, Google AI Overviews. The user gets a ready conclusion plus a list of sources it was built from. Becoming one of those sources is the new job of promotion, and it isn't solved the way classic SEO is.
This article is a working method: what GEO and AEO are, how AI engines decide whom to cite, and what to do on your site this very week.
What GEO and AEO are
GEO (generative engine optimization) is optimization for generative engines: making sure the model finds your content, selects it as a source and cites it in the answer. AEO (answer engine optimization) is the broader term: optimization for answer engines, including voice assistants and direct-answer boxes. In practice the boundary is blurred and you work on both at once.
The key difference from SEO: a search engine ranks pages, while an AI engine selects passages. A whole page earns a top ranking, but what gets into an AI answer is a specific paragraph — provided it is self-contained, precise and easy to extract.
How AI engines pick their sources
Almost all of them work the same way: the query is reformulated, a search runs against their own or a partner index, relevant fragments are extracted from the found pages, and the model synthesizes an answer with references. So getting into the answer means passing three filters:
- Indexing — the crawler must be able to reach your content freely.
- Passage selection — the fragment must answer the question directly and completely.
- Citation — the model must “trust” the source: entities, authority, consistent data.
The GEO/AEO checklist: 8 steps
- Open access for AI crawlers in robots.txt — GPTBot, ClaudeBot, PerplexityBot, Google-Extended and others — and check your CDN or firewall isn't blocking them.
- Add llms.txt and llms-full.txt — a curated, citation-ready index of your content.
- Write in citable passages: one question — one 40–80-word paragraph with the direct answer in the first sentence.
- Add FAQ blocks with FAQPage markup: question-and-answer is the format models extract best.
- Mark up entities in JSON-LD: Organization, Service, Article, BreadcrumbList; names and descriptions must match across every source.
- Publish specifics, not slogans: numbers, timelines, checklists, comparisons — models cite facts.
- Link language versions with hreflang: engines answer in the user's language and pick sources in it too.
- Keep content and dateModified fresh: recency is a signal for both search and AI.
How to measure results
There are no “rankings” here, but the result is measurable. Regularly ask the engines your key questions and record whom they cite. Watch referral traffic from chatgpt.com and perplexity.ai in your analytics. Track branded-query growth: people often see you in an AI answer first and come back later through search.
Common mistakes
- Blocking AI bots “just in case” — you're simply absent from the source pool.
- Wall-of-text content with no structure: the model won't find the passage even if the answer is in there.
- Marketing language instead of facts — it doesn't get cited.
- Expecting instant results: engines re-crawl content over weeks.
GEO/AEO doesn't replace SEO — it extends it: the same principle of being the best answer, except the answer is now assembled by a machine. Make your content easy to cite, and you'll start being found where your audience actually asks its questions.