What is AEO? Answer engine optimisation, explained

What is answer engine optimisation?

AEO is the practice of structuring your site so that answer engines — ChatGPT, Perplexity, Claude, Google's AI Overviews, and the growing list behind them — can read it, understand what you do, and cite you when someone asks a question you can answer. Traditional search optimises for a ranked list of blue links a person clicks. AEO optimises for the synthesised answer that sits above the links, or replaces them entirely.

The shift matters because the unit of visibility has changed. In classic search, being on page one is the goal, and the click is the reward. In an answer engine, there may be no list and no click. There is one answer, assembled from a handful of sources the model decided it could trust, with a few citations attached. You are either in that answer or you are invisible, and being invisible is quiet — no ranking drop to notice, no traffic cliff to investigate, just a slow absence from conversations you never see.

AEO vs SEO: what actually changed

AEO is not a replacement for SEO. It is a second surface that sits on top of the same foundations. Most of what makes a page rank well also makes it citable, which is why teams with strong technical SEO tend to get a head start. But the goals diverge in ways that change what you prioritise.

The practical consequence: a page that is beautiful to a human and opaque to a parser can rank and still never be cited. If your value proposition lives inside an image, your pricing is rendered by JavaScript a crawler never runs, and your key claims are implied rather than stated, a model has nothing clean to lift. It will either paraphrase you vaguely or quote a competitor who wrote it down plainly.

How AI engines decide what to cite

No two answer engines work identically, and none of them publish their exact criteria. But the mechanics rhyme, because they are all solving the same problem: given a question, retrieve a small set of trustworthy sources, and generate an answer that can point back at them. Four things repeatedly decide whether you make the cut.

Retrievability

Before a model can cite you it has to be able to fetch and parse you. That means AI crawlers must be allowed in robots.txt, your important content must be present in the HTML rather than assembled by client-side script the crawler never executes, and the page must return cleanly rather than behind an interstitial or a login. This is the step most sites fail without realising, because the block is usually accidental.

Extractability

Once fetched, the page has to yield facts. Semantic HTML, a sane heading hierarchy, and plainly stated claims let a model pull a clean sentence to quote. A page built from nested divs with visual-only hierarchy reads to a parser as one long undifferentiated block, and what gets quoted is whatever happened to fall near the top.

Corroboration

Models prefer claims they can see echoed elsewhere. If your homepage, your structured data, your llms.txt and a couple of third-party mentions all agree on what you are and what you cost, that consistency reads as reliability. If they contradict each other, the model hedges — and a hedged answer rarely names you.

Specificity

Generic copy is un-citable by design. "Revolutionise your workflow" answers no question a person actually types. A page that states who it is for, what it does, what it costs, and what it is not gives an engine something to match against a real query. Specificity is the single cheapest lever most sites have and the one they use least.

The signals an AEO audit checks

Presence checks are the deterministic core of AEO — either the file is there and correct or it is not, with nothing to argue about. The free audit runs seven of them against any URL and scores the result. None of them measure whether an engine actually cites you, which no presence check can; they measure whether an engine is able to.

What is llms.txt?

llms.txt is a plain-text file at the root of your domain — the same place robots.txt lives — that tells a language model what your site is, which pages matter, and how to describe you. It is the one artifact where you get to write your own summary rather than hoping a good one is inferred from your navigation. The format is deliberately simple: a title, a short description, and curated links under markdown headings.

# Acme Analytics

> Product analytics for B2B SaaS teams. Self-serve, privacy-first,
> deployed in an afternoon.

## Key pages
- [Pricing](https://acme.com/pricing): plans from $0, usage-based above
- [Docs](https://acme.com/docs): setup, SDKs, and the query API
- [Security](https://acme.com/security): SOC 2, data residency, retention

Support is still uneven — no engine treats llms.txt as authoritative yet — which is exactly why it is cheap insurance. It takes an afternoon, has no downside, and hands the model a clean brief instead of forcing it to reconstruct one from your header. Write it the way you would brief a new salesperson who will never meet you: what you sell, who for, what you are not, and the five URLs that matter.

How to improve your AEO

The work divides cleanly into three phases, cheapest and highest-leverage first. Most sites can close the first phase in a week.

Phase 1 — Let the machines in and give them facts

• Check robots.txt for accidental blocks of GPTBot, ClaudeBot, PerplexityBot and Google-Extended, and decide access deliberately rather than by inherited template.
• Add JSON-LD for the types you actually are: Organization and WebSite at minimum, then Product, SoftwareApplication, FAQPage or Article as they apply.
• Publish an llms.txt with your own summary and your five most important URLs.

Phase 2 — Make the page extractable

• Fix the heading hierarchy: one h1, headings that descend in order, real header/nav/main/article/footer landmarks.
• Move key claims — what you do, who for, what it costs — out of images and JavaScript and into plain, crawlable HTML.
• Complete the meta and Open Graph set so a shared link renders with a real title, description and image.

Phase 3 — Write like a question is being asked

• Add an honest FAQ in customer words — "how much does it cost", not "pricing philosophy" — and mark it up as FAQPage.
• State the specifics competitors leave vague: real numbers, real constraints, real comparisons.
• Keep your sitemap current so new pages are found and dead ones are not advertised.

Frequently asked questions about AEO