GEO: optimising your site to be cited by ChatGPT, Perplexity and Google AI

GEO (Generative Engine Optimization) is the practice of structuring and writing your site so it gets cited as a source by generative engines — ChatGPT, Perplexity, Google AI Overviews, Gemini or Claude. In concrete terms: you no longer aim only to appear in a list of links, but to become one of the few sources an AI summarises and attributes in its answer. It's a discipline that complements classic SEO — not a replacement for it.

GEO vs classic SEO: what really changes

SEO aims to rank a page among results: the goal is the click. GEO aims to be extracted and cited inside a synthesised answer: the goal is the mention, with or without a click. Both share a common base — useful content, fast site, clean indexing — but the end goal differs, and so does the way you write and structure.

• SEO: optimise for a ranking algorithm that orders links.
• GEO: optimise for a model that reads, summarises and attributes information to a few sources.
• Common ground: without proper crawling and indexing, neither works.

How generative engines choose their sources

The exact mechanics remain partly opaque and evolve fast, but several constants stand out. To answer, a generative engine relies either on content already seen during training, or — for Perplexity, AI Overviews or ChatGPT in search mode — on real-time retrieval of indexed pages. In both cases, it favours passages it can cite safely: clear, verifiable, attributable to a credible entity.

• Relevance: the passage answers the exact question asked.
• Extractable clarity: the information stands on its own in a sentence or block, with no missing context.
• Source reliability: identified entity, demonstrated expertise, dated and signed content.
• Consistency: the same information is confirmed elsewhere on the web (a consensus signal).

First concrete step: allow the AI crawlers

A generative engine can only cite what it's allowed to read. Many sites unintentionally block AI crawlers in their robots.txt, or allow nothing explicitly. To be eligible for real-time retrieval and indexing, you have to decide — deliberately — which agents to allow.

• GPTBot — OpenAI's crawler (training and retrieval for ChatGPT).
• OAI-SearchBot — retrieval for ChatGPT search.
• ClaudeBot — Anthropic's crawler for Claude.
• PerplexityBot — Perplexity's crawler.
• Google-Extended — controls use by Gemini and Google's AI features (separate from Googlebot).

Write “answer-first” and structure for extraction

The golden rule of GEO: give the answer first, the context second. An AI extracts crisp statements, not introductions that circle the topic. Every section benefits from opening with a self-contained sentence that directly answers the implicit question in its heading, then expanding.

1. 01Direct answer in the first sentence of each section, understandable out of context.
2. 02Explicit headings phrased as questions or clear statements (h2, h3).
3. 03Lists and tables to break down enumerations, comparisons and steps.
4. 04FAQ at the bottom: one question, one concise, self-contained answer.
5. 05Definitions and figures highlighted, easy to isolate and cite.

An AI doesn't cite a page: it cites a passage. If that passage needs the previous paragraph to make sense, it won't be picked up.

Structured data and strengthening the entity

Schema.org structured data helps machines understand what is what on your page: an article, its author, its date, an FAQ, an organisation. It doesn't guarantee a citation, but it removes ambiguity and reinforces attribution. The most useful markups for GEO are Article, FAQPage, Organization and Person (for the author).

Beyond markup, it's your entity you need to consolidate — the E-E-A-T framework (Experience, Expertise, Authoritativeness, Trustworthiness). An AI more readily picks up a source it recognises as real and competent: identified author, company with legal information, expertise demonstrated through signed and dated content, consistent presence across the web. This technical and editorial base also depends on your framework — we explain it in our article on Next.js and SEO.

GEO is part of a broader transformation of search: if you want the full picture, read SEO and AI in 2026, our pillar guide on the topic.

An actionable GEO checklist

Here's a concrete checklist to apply page by page. It doesn't replace a strategy, but it covers the fundamentals that separate an ignored page from a cited one.

• Your robots.txt explicitly allows the AI crawlers you want (GPTBot, ClaudeBot, PerplexityBot, Google-Extended...).
• Each page answers one clear intent, from the very first paragraph.
• Headings are explicit and structured in a logical hierarchy (h1 > h2 > h3).
• Key information sits in lists, tables or self-contained sentences that are easy to extract.
• An FAQ covers users' concrete questions, with short answers.
• Structured data for Article, FAQPage and Organization is in place and valid.
• The author and date are visible; the entity is credible (legal info, web consistency).
• The site is fast and server- or statically rendered: content a crawler can't read doesn't exist for it.

How to measure your AI citations

GEO forces you to measure differently. Classic analytics tools don't see a mention inside a ChatGPT answer. Measurement combines several approaches — still imperfect, but already useful for tracking a trend.

• Recurring manual tests: ask ChatGPT, Perplexity and Gemini the target questions of your business and note whether you're cited.
• Referral traffic: watch for visits from chatgpt.com, perplexity.ai or AI Overviews in your analytics.
• GEO tracking tools: emerging platforms monitor brand mentions in AI answers — useful to follow a trend, to be read with caution.
• Server logs: confirm that AI crawlers (GPTBot, PerplexityBot...) actually hit your pages.