How to get your website cited by ChatGPT, Perplexity and generative AI

Understand how AI engines choose their sources

Answer engines combine two mechanisms. First, the model's internal knowledge, built during training: being part of it requires a lasting footprint on the web (press, directories, content picked up by other sites). Second — and above all — live retrieval (RAG): ChatGPT Search, Perplexity or Gemini query the web at question time, read a handful of pages and compose the answer while citing their sources.

The practical consequence: to be cited, your page must be accessible to AI crawlers, surface in the underlying searches (often Bing for ChatGPT), and contain passages that are easy to extract and attribute. Each step below addresses one of these links in the chain.

Step 1: open your website to AI crawlers

First reflex, often forgotten: check your robots.txt. A block inherited from a plugin or an anti-bot protection can make you completely invisible. Explicitly allow the main crawlers:

• GPTBot and OAI-SearchBot — OpenAI's crawlers, for model training and for ChatGPT Search.
• ClaudeBot — Anthropic's crawler, used for Claude.
• PerplexityBot — Perplexity's crawler, the engine that systematically cites its sources.
• Google-Extended — Google's access for Gemini and generative AI features.

Also make sure your key pages load without requiring JavaScript: most AI crawlers read the raw HTML without executing your scripts.

Step 2: publish an llms.txt file

llms.txt is a Markdown file placed at the root of your website (example: yoursite.com/llms.txt) that presents your business and lists your important pages with a short description. It is the equivalent of a sitemap designed for large language models: it saves them from guessing who you are and guides them to your best pages. Quick and simple to create, it is one of the eight criteria audited by GeoGenee.

Step 3: make your content citable

Models cite passages, not websites. A citable passage is short, factual and self-contained: it answers the question from the very first sentence, without depending on the previous paragraph.

• Answer first, elaborate later — under each H2 phrased as a question, give the answer in one or two sentences before going into detail.
• Provide figures and dated facts — statistics, studies and concrete examples are massively picked up by generative engines.
• Add an FAQ — your customers' real questions, with complete, self-contained answers.
• Show your dates — visible publication and update dates: search-augmented engines favour recent content.

Step 4: mark up with structured data

schema.org markup in JSON-LD turns your prose into machine-usable facts: Organization for your identity, FAQPage for your questions and answers, Article for your guides, Product or LocalBusiness depending on your business. These schemas remove ambiguity — a model that knows precisely who you are and what you offer cites you with much more confidence.

Step 5: build the authority AI engines verify

AI engines cross-check their sources. A brand cited by the trade press, present in reference directories and comparison sites, with consistent customer reviews, becomes a "safe" answer the model dares to recommend.

• Target the sources AI already cites — ask your target questions to Perplexity and note the cited domains (comparison sites, directories, media): get a presence there.
• Keep everything consistent — same name, description and contact details everywhere — website, LinkedIn, directories, Google Business Profile.
• Publish expert content — studies, guides and original data other websites pick up: the strongest authority signal there is.

Step 6: measure your citations, engine by engine

You cannot steer without measuring: visibility varies strongly from one engine to another and changes every week. Define 10 to 30 questions your customers actually ask and track, for each one: are you cited, at what position, with what sentiment, against which competitors?

GeoGenee automates this tracking across 10 engines (ChatGPT, Claude, Gemini, Perplexity, Mistral, Grok…), computes your share of voice against your competitors and audits your website on the 8 GEO criteria — with a prioritised action plan.

Frequently asked questions