2026-08-03 · Jack Stovell
Generative engine optimisation (GEO), explained
Generative engine optimisation (GEO) is the practice of making your content something AI-generated answers retrieve, quote and cite — in ChatGPT, Perplexity, Google's AI Overviews and their peers. It stands on the same foundations as SEO: crawlable pages, credible content, a clear identity. What changes is the unit of competition — answers assembled from passages, rather than ranked lists of links.
What is generative engine optimisation?
GEO is optimising for systems that answer questions instead of listing results. When an assistant answers, it retrieves a small set of sources and composes a response from them; GEO is the work of being in that set, and being quotable once you are. The term was popularised by a 2023 research paper, and the industry has since attached several names to the same idea — AI SEO, answer engine optimisation, LLM SEO. They all describe the same shift.
It is a young discipline, and worth saying so plainly: the evidence base is thin, the engines change monthly, and nobody outside these companies knows the ranking internals.
GEO vs SEO: what actually changes?
The foundations do not change: if a system cannot crawl your pages, or has no reason to treat them as authoritative, neither SEO nor GEO can start. On top of that shared base, five things genuinely shift:
- Answers, not links. The surface is a composed answer; the win condition is being among its sources, not earning the click.
- Passage-level extraction. Engines lift the paragraph that answers the question, not the page as a whole. Direct, self-contained passages under question-shaped headings get quoted; answers buried mid-essay don't.
- Entity identity matters more. A model has to be certain which person or company it is talking about before it says anything firm. Structured data, consistent naming and sameAs links do that work.
- Being quotable beats being clicked. Plain claims with dates and numbers a model can repeat accurately outperform clever copy that only works in context.
- Zero-click is the norm. Many answers end without a visit, so the citation is the win — and measurement shifts from traffic to mentions.
There is also a plumbing difference worth knowing: Googlebot renders JavaScript, most AI crawlers do not. A site can be indexed by Google yet invisible to AI.
How do AI Overviews fit in?
Google's AI Overviews are generated from Google's ordinary Search index — there is no separate crawler to court and no submission path, so ordinary SEO is what feeds them. If your pages are indexed and rank for a query, they are in the pool Overviews draw from.
One control is worth stating precisely, because it is widely muddled: Google-Extended, the robots.txt token, governs whether your content is used to train Google's Gemini models. As Google documents it, it is not the switch that decides inclusion in Search — and AI Overviews are part of Search. Blocking Google-Extended is a training decision, not an Overviews opt-out. How Overviews select among indexed pages beyond that is not fully documented, and any confident claim about it deserves suspicion.
What does GEO work actually look like?
In practice, GEO work is unglamorous: remove the reasons you can't be cited, then make your best pages easy to quote.
- Fix crawl access: your robots.txt stance per AI user agent, CDN bot-blocking defaults, login walls.
- Serve real HTML: server-rendered content that survives a fetch with no JavaScript.
- Answer directly: question-phrased headings whose first sentence is the answer.
- Establish identity: Person or Organization JSON-LD, sameAs links, one consistent name and description everywhere.
- Keep receipts: dates on claims, and sources you would be happy to see quoted back at you.
- Measure honestly: ask the assistants your target questions each month and record what they cite.
And the honest close: GEO is young enough that anyone selling guaranteed AI rankings is guessing. We publish our own results — including the tracked queries where models do not yet cite hilyt profiles — on our public scoreboard, because a field this new needs receipts more than promises. Being AI-readable removes disqualifiers; retrieval, ranking and model behaviour decide the rest. You can see where you stand today with the free AI visibility checker.
Check your AI visibility — free
Related guides
- How AI assistants decide which sources to cite
- How to get cited by Perplexity
- How to get seen in ChatGPT — the mechanics, not the myths