What is Generative Engine Optimisation (GEO)?

GEO in one sentence

Generative Engine Optimisation is the work of becoming a source an AI answer engine can confidently retrieve and cite. It does not replace search engine optimisation. It adds a passage-level retrieval and citation problem on top of the familiar crawl, index and ranking problem.

Google's own guidance says that the same foundational SEO practices remain relevant for AI Overviews and AI Mode, and that a page must be indexed and eligible for a snippet before it can appear as a supporting link. That makes classical search visibility the entry ticket, not the entire game.

GEO versus SEO

What changed when search began writing answers

A conventional result page retrieves and orders documents. An AI search experience can fan a question out into related searches, retrieve passages from several sources, rerank them, synthesize an answer and attach citations. Google publicly describes query fan-out for AI Overviews and AI Mode. OpenAI likewise explains that ChatGPT search can rewrite a prompt into one or more targeted queries.

The consequence is practical: a page can rank well yet fail to contribute a passage, while a tightly written section from another page can be retrieved for one sub-question and become the cited source.

The five gates between publication and citation

1. Access: the search or answer engine can fetch the URL without a robots, firewall, challenge or authentication block.
2. Indexing: the page is canonical, indexable and eligible to appear with a snippet.
3. Retrieval: the page contains language that matches the question and its likely sub-questions.
4. Selection: the retrieved passage is specific, useful, supported and clearer than competing passages.
5. Attribution: the system can associate the claim with a stable URL and a recognisable author, organisation or entity.

What good GEO work looks like

• Answer the section's question in its first one or two sentences.
• Give each heading one clear job and use the vocabulary a searcher would use.
• State the subject explicitly instead of relying on pronouns and surrounding context.
• Add evidence, dates, methods and source links where a claim depends on them.
• Keep facts consistent across visible copy, structured data, feeds, profiles and third-party listings.
• Create original assets: definitions, comparisons, data, procedures, diagrams and clearly scoped expert judgement.
• Measure citations across a fixed prompt set instead of treating one anecdotal answer as a trend.

What GEO is not

GEO is not a special schema type, a magic AI file or permission to manufacture large volumes of near-duplicate copy. Structured data can reduce ambiguity when it matches the visible page, but Google states that there is no special schema required for its AI features. An llms.txt file can be a useful voluntary summary for systems that choose to read it; it does not replace crawlable HTML, sitemaps, internal links or indexation.

A practical first week

1. Confirm the canonical domain, robots rules, sitemap and index status in Google Search Console and Bing Webmaster Tools.
2. Choose the ten questions that matter most commercially and informationally.
3. Create or improve one definitive page for each distinct intent; do not force unrelated questions into one page.
4. Rewrite each important section answer-first, then add proof and a useful next step.
5. Record a baseline across the same prompts in the assistants your audience uses.
6. Submit changed URLs through the relevant webmaster tools and IndexNow, then measure over time.