AEO: how to write pages AI assistants quote

Answer engine optimisation (AEO) is shaping your pages so AI assistants such as ChatGPT, Perplexity and Google’s AI Overviews can find, understand and quote them. In practice it’s mostly well-written pages on a technically sound site: the answer first, one question per section, specific facts with sources, and nothing hidden from crawlers. Google says there are “no additional requirements” for its AI Overviews and AI Mode beyond that (Google).

Here’s how we set up our own site, what an independent SEO and AEO review in October 2026 told us to change, and the tools we use to see whether it’s working.

What is answer engine optimisation?

An answer engine gives one written answer instead of a list of links, usually citing a few sources. ChatGPT, Perplexity, Claude, Copilot, and Google’s AI Overviews and AI Mode all work this way. AEO is the work of becoming one of those sources: pages that answer real questions directly, facts that are specific and sourced, and a site that crawlers can read.

You’ll also see it called GEO, generative engine optimisation, or LLMO. They describe the same job (AEO vs GEO vs SEO). It isn’t separate from SEO: the assistants mostly cite pages that search engines can already find and trust, so AEO builds on SEO rather than replacing it.

	SEO	AEO
The goal	A page ranking in search results	Your page quoted or cited in an AI answer
What people see	A list of links	A written answer, with a few sources
What it rests on	A crawlable, fast site with clear pages	The same, plus direct answers and specific, sourced facts
How you measure it	Rankings, clicks and Search Console	Citations, mentions and visits from AI assistants

What we built into our own site

Answers first, and FAQs where the questions come up. Every service page leads with what the service is, and ends with the questions people ask about it before they start. There’s no single FAQ page: each answer sits on the page it belongs to, and opens with the direct answer in its first sentence.

One source for the questions and their markup. The FAQ markup on each page is generated from the same list as the questions you see, so the two can’t drift apart. Google says structured data should match the visible text on the page, and this is the simplest way we know to keep it that way (Google).

Everything in the HTML. Page copy and FAQ answers are in the page the server sends, so nothing depends on a crawler running scripts or opening an accordion. The review checked this across every page and found it held.

Structured data that describes what’s there. The site describes the studio as an organisation and a professional service, each service with its breadcrumbs and FAQs, the founder as a person, the work as a list, and each journal post as an article with its author. Google says there’s no special markup for AI features, so we treat it as description, not a shortcut.

Crawlers let in on purpose. Our robots.txt allows the search crawlers and the AI crawlers, including the ones that collect content for training. That’s a choice, not a requirement. Search and training are separate: OpenAI’s OAI-SearchBot and Anthropic’s Claude-SearchBot fetch pages for search answers, while GPTBot, ClaudeBot and Google-Extended relate to training, so a site can block training and stay visible in AI search (OpenAI, Anthropic, Google). We want assistants to describe us accurately, so we let them all in.

An llms.txt file, kept modest. We keep a plain summary of the site at /llms.txt, listing the services, case studies and journal posts. Google says Search ignores these files and they “will neither harm nor help” (Google). Some other tools read them, so it’s cheap to keep accurate, but we don’t count on it.

What the review told us to change

The biggest fixes the review asked for weren’t technical. They were about accuracy and evidence:

• Our own SEO page overclaimed. It said robots rules and an llms.txt file would help AI crawlers “get your key facts right”. Crawler access isn’t the same as being quoted, so it now describes crawler checks and clear, linked content, with llms.txt as an optional extra (SEO & AI search).
• A migration answer promised too much. We said a well-planned move wouldn’t lose rankings. It now says a careful migration keeps the risk low, but no one can promise rankings won’t move.
• The practical questions were missing. Buyers want to know what affects the cost, how long a project takes, what they need to supply and who owns the site afterwards. Those answers are now on the Development page.
• Our strongest page was hard to find. The healthcare page was linked from one other page. It’s now linked from the home page, the services that feed into it, and back again.

The review was just as clear about what wasn’t the gap: more schema, or another llms.txt project. Its advice was to treat AI search as evidence and reputation, not a checklist.

What Google changed while we were doing it

In May 2026 Google stopped showing FAQ rich results in Search, and later removed their documentation (Google). We kept the FAQ markup anyway. It still describes the questions on the page accurately, and the questions are there for readers first. If a tactic only works because of a special search feature, it was never the point.

The same habits on client sites

The pattern holds on the sites we build:

• Altus has FAQ markup on its three product pages, alongside structured data for the organisation, its software, articles and breadcrumbs.
• Fresh Clinics stores more than 90 FAQs once in its CMS, written with its SEO agency, and shows each on the pages it belongs to.
• Generate describes the organisation, the Summit as an event and its FAQs in structured data, with an llms.txt that says what Generate is.

One of Altus’s FAQ sections. Its three product pages pair questions like these with FAQ markup.

Being cited isn’t being recommended

A citation is your page linked as the source for a fact. A recommendation is your name on a shortlist when someone asks for the best options. Clear pages help with the first. The second depends far more on what others say about you: reviews, directories, press and genuine discussion. Google’s guide notes its AI features draw on what’s said in blogs, videos and forums, and that inauthentic mentions don’t help much (Google). For us that means real Google reviews, an accurate Webflow partner profile, and clients linking to the work we’ve done together.

How we measure it

Answers change from one run to the next, so we look at trends rather than single results:

1. A fixed set of prompts, the questions our clients ask, run monthly in the assistants they use.
2. Three results recorded separately: cited, mentioned and recommended. They move for different reasons.
3. Search Console, where AI Overviews and AI Mode count towards the Performance report (Google).
4. Referrals. Our analytics records a visit_from_ai event when a visit comes from ChatGPT, Perplexity, Claude, Gemini or Copilot. It undercounts, because some apps send no referrer and people who come back later look like any other visit, so we treat it as a floor.

The AEO tools we use

No single tool shows whether AI assistants quote you, so we combine a few:

Tool	What it tells us
Google Search Console	Searches, clicks and indexing, with AI Overviews and AI Mode counted in the Performance report
Bing Webmaster Tools	The same for Bing; our site also uses IndexNow to tell Bing when a page is added or changed
PostHog	Which pages visitors from AI assistants land on, from our visit_from_ai event
DataForSEO	How often people search for a question, and how hard it is to rank for, so each post answers a question people actually ask
The assistants themselves	Our fixed prompt set, run by hand each month

Schema validators, such as the Schema Markup Validator, are worth a check after any change to structured data. On Webflow sites, the Audit panel also flags missing titles, descriptions, alt text and schema markup (Webflow SEO).

Questions we get asked

Do we need to write differently for AI assistants?

No. Write clearly for the people you want to reach: answer first, one question per section, specific facts with sources. Google says its AI features need no special optimisation, and the same habits help readers and every other assistant.

Should we block GPTBot or ClaudeBot?

That’s a choice about training, not visibility. GPTBot and ClaudeBot collect content for model training, while OAI-SearchBot and Claude-SearchBot serve search answers. You can block the training crawlers and still allow the search ones. We allow both, because we want to be described accurately.

Is FAQ markup still worth adding?

Yes, as an accurate description of questions that are already on the page. Google stopped showing FAQ rich results in May 2026, so it no longer earns a special result, but it costs little when it’s generated from the same content readers see.

Can you guarantee we’ll be cited by ChatGPT or AI Overviews?

No one can. We can remove the technical barriers, structure the content and help it say clearly what you do, which is what these systems rely on. Being recommended also depends on reviews and coverage outside your site.

What are the best AEO tools?

There isn’t one tool that shows everything. We use Google Search Console and Bing Webmaster Tools for search and indexing, analytics that records visits from AI assistants, a keyword tool to choose the questions to answer, and a fixed set of prompts run by hand in the assistants themselves each month.

What is AEO vs SEO?

SEO is getting pages to rank in search results. AEO, answer engine optimisation, is getting your content quoted in AI answers, such as Google’s AI Overviews, ChatGPT and Perplexity. They share the same foundations: clear pages, answers first, sources and a technically sound site.

Is SEO dead now with AI?

No. Google says its AI Overviews and AI Mode need nothing beyond good SEO, and AI assistants cite pages that search engines can already find and trust. Good SEO is still the foundation; AEO adds clear answers and structure on top.

If you’d like your site checked for what search engines and AI crawlers can see, that’s where our SEO & AI search work starts. Tell us about your site.

This guide is part of Webflow for marketing teams, which brings together everything we’ve written on choosing, building and running a Webflow site.

Sources

• Google Search Central, AI features and your website, read 9 October 2026
• Google Search Central, Optimizing your website for generative AI features on Google Search, read 9 October 2026
• Google Search Central, Documentation updates: the FAQ rich result deprecation (May 2026) and the llms.txt guidance (June 2026)
• Google, Google’s common crawlers, read 9 October 2026
• OpenAI, Overview of OpenAI crawlers, read 9 October 2026
• Anthropic, Does Anthropic crawl data from the web?, read 9 October 2026