AI SEO: generative engine optimisation in plain English

AI SEO, also called generative engine optimisation (GEO) or answer engine optimisation (AEO), is the work of making your business easy for AI search tools such as Google’s AI Overviews, ChatGPT search, Perplexity and Claude to find, read and cite. Most of what the products themselves document is ordinary SEO: let their crawlers in, publish pages that answer real questions clearly, and keep your business details accurate everywhere.

Much of what is sold beyond that is guesswork. This article separates what Google, OpenAI, Perplexity and Anthropic have published from what people are inferring, so you can decide what is worth paying for. If you want a practical checklist first, our guides to getting recommended by ChatGPT and getting into Google AI Overviews cover the basics.

What AI SEO and generative engine optimisation mean

An AI search tool answers a question in its own words and, usually, links to the pages it drew on. AI SEO is about being one of those pages, or one of the businesses named in the answer.

The term “generative engine optimisation” comes from a research paper, GEO: Generative Engine Optimization, first posted in November 2023 by Pranjal Aggarwal and five co-authors and accepted to the KDD 2024 conference. The authors built a test set of queries, called GEO-bench, and reported that some changes to a page’s wording could raise its visibility in generated answers by up to 40% in their tests. That is a result on a research benchmark. It is not a measurement of ChatGPT or Google, and the authors themselves found that what worked varied by subject.

“Answer engine optimisation” is an older, looser term for the same idea. Google’s own position is that these labels describe SEO. Its guide to optimising for generative AI features mentions AEO and GEO by name, says that optimising for generative AI search is still SEO, and suggests judging outside AEO and GEO advice the same way you would judge any outside SEO advice.

What the AI search products themselves document

Each of the main products publishes how its crawlers work. None publishes how it chooses which sources to cite. Here is what each says, as at November 2026. Check each page for changes.

The sources: Google’s page on AI features and your website, OpenAI’s crawler overview, Perplexity’s crawler page and Anthropic’s crawler help article.

Two details matter for owners. OpenAI and Perplexity both say a robots.txt change can take up to about a day to reach their search systems. And each product treats search and training separately, so you can let a tool cite your pages while telling it not to use them for training.

What Google says you do not need for AI SEO

Google has gone further than the others and listed things it says make no difference to its AI features. According to its generative AI optimisation guide and its AI features page:

• No special files. You do not need machine-readable files, AI text files or Markdown versions of pages. Google Search does not use llms.txt; our guide to llms.txt covers what it does elsewhere.
• No special schema. There is no AI-specific structured data. Structured data still helps normal search features.
• No chunking. You do not need to cut content into small pieces for AI to understand it.
• No ideal length. Google says there is no ideal page length and you should write for your audience.
• No rewriting for AI. Its systems understand synonyms, so you do not need a page for every variation of a question.

Google also warns against two popular tactics. Creating pages for every related query its systems might run, which it calls query fan-out, counts as scaled content abuse under its spam policies. And it says inauthentic “mentions” planted across the web are not as helpful as they seem. It adds that no third-party tool has access to its internal ranking or AI systems.

What is guesswork in AI SEO

A lot of AI SEO advice is reasonable inference, some is guesswork, and some is sold as certainty. Be clear which is which before you pay for it.

• How ChatGPT, Perplexity or Claude choose which businesses to name. None of them publishes this. Anyone who says they know the formula is inferring it from testing.
• “AI ranking” reports. Tracking tools ask AI tools a set of questions and record the answers. That is useful as a sample, but answers change from one run to the next and by person, so a single report is a snapshot.
• Lab tactics. The GEO paper tested changes such as adding statistics and quotations on a benchmark. Whether the same changes help a local plumber’s service page in a live product is unknown.
• Brand mentions. It is plausible that being described on many independent sites helps, because AI tools cite other sites. No product documents how much weight it gives them, and Google says fake mentions do not help.

Nobody can promise that a tactic will get your business recommended. The documented work raises your chances; it does not guarantee a place in any answer.

AI SEO work that is worth doing

The work that is documented, and that also helps ordinary search, is the same list whichever tool your customers use.

1. Check who can read your site. Look at yourdomain.com/robots.txt and your firewall or security plugin. Make sure Googlebot, OAI-SearchBot, PerplexityBot and Claude-SearchBot are not blocked by accident, then decide on the training crawlers separately.
2. Answer first. Put the direct answer to each customer question in the first sentence or two of the page, then the detail. AI tools quote openings.
3. Add what only you know. Google recommends unique, first-hand content. Your prices or what affects them, your process, your service areas and the problems you see every week are things a summary of other websites cannot copy.
4. Keep business details identical. Google says AI responses can include local business information and that a Google Business Profile helps. Use the same name, phone and address on your site, your profile and directories.
5. Use structured data for the normal reasons. Our guide to schema markup explains what is worth adding.
6. Earn genuine reviews and mentions. Real customers describing real jobs give every tool something to work with.

If you write with AI tools yourself, read whether Google penalises AI content before publishing a large batch of pages.

How to measure AI search visibility

There is no single AI ranking report, so use three sources together.

• Google Search Console. Google says visits from AI Overviews and AI Mode are counted in the Performance report under the Web search type, and its newer optimisation guide points to a Generative AI performance report as well.
• Your analytics. Look for visits referred from chatgpt.com, perplexity.ai, claude.ai and similar sites, and for enquiries that follow them.
• A monthly question check. Ask ten real customer questions in each tool, record whether you are named and which sources are cited, and watch the pattern over three months.

The number that matters is enquiries. A mention that never turns into a call is a vanity figure.

Questions to ask anyone selling AI SEO

• Which published documentation from Google, OpenAI, Perplexity or Anthropic supports this work?
• Which pages will you change, and what will be different on each?
• Will you generate pages in bulk? If so, how does that fit with Google’s scaled content abuse policy?
• How will you report results, and how will we see enquiries rather than mentions?
• Are you promising a placement? Google’s own advice on hiring an SEO says nobody can guarantee a top ranking, and the same applies to AI answers.

CentriWeb handles AI search visibility as part of our SEO and marketing work: crawler and site checks, answer-first service pages and accurate business listings, with reporting tied to enquiries. Where your site already does this well, we will say so.