Answer Engine Optimisation: How to Get Cited by ChatGPT, Claude and Perplexity
Search is splitting into two behaviours: finding links and receiving answers. Answer engine optimisation is how you win the second without abandoning the first.
For twenty years the objective of search visibility was a position on a results page. That objective is now only half the job. A growing share of commercial research begins inside an assistant — a question typed into ChatGPT, a comparison asked of Claude, a supplier shortlist requested from Perplexity — and ends with a synthesised answer that names two or three sources.
Answer engine optimisation is the discipline of being one of those sources. It is not a replacement for search engine optimisation; it is what SEO becomes when the retrieval unit is a passage and the reward is a citation.
How answer engines actually decide what to cite
The mechanics differ between products, but the pipeline is consistent enough to optimise against. A user question is interpreted and often decomposed into sub-questions. Each sub-question triggers retrieval — from a search index, a vector store, a live crawl, or all three. Candidate passages are re-ranked. A subset is passed to the model as grounding context, and the answer is generated with attribution to whatever survived.
Four properties determine survival:
1. Retrievability
Content that requires JavaScript execution, sits behind an interstitial, or lives only inside an accordion that never renders server-side is frequently invisible to the crawlers feeding these systems. Google documents its own AI features and crawler controls separately from traditional search, and other providers publish distinct user agents (see sources).
2. Extractability
Assistants quote passages, not pages. A self-contained paragraph that answers a specific question — subject named, claim stated, qualification included — can be lifted intact. A brilliant argument that unfolds over nine paragraphs cannot.
3. Corroboration
Models weight agreement. A claim repeated consistently across your own site, your structured data, and independent reputable sources is treated as more reliable than a novel assertion, however true.
4. Attributable authority
Named authors with verifiable credentials, an unambiguous publishing organisation, dates, and citations of primary sources are the machine-readable proxies for the trust signals Google describes as experience, expertise, authoritativeness and trustworthiness.
The passage-level content pattern that works
Rewrite important pages around this structure and citation rates change measurably.
- Question-shaped heading. Use the phrasing a human would type or speak, not a keyword fragment.
- Direct answer in the first two sentences. Forty to sixty words, complete on its own, no pronouns referring backwards.
- Then the reasoning. Evidence, trade-offs, numbers, caveats.
- Named entities restated. Say "the European Accessibility Act", not "the Act", because the passage may be read with no surrounding context.
- Concrete specifics. Dates, versions, thresholds, prices and measurable claims are what distinguishes a citable source from generic commentary.
Technical foundations you cannot skip
| Signal | Why it matters for AEO | Practical action |
|---|---|---|
| Server-rendered HTML | Many AI crawlers do not execute JavaScript reliably | Pre-render or SSR all content pages |
| Connected schema graph | Disambiguates your organisation, authors and services | Publish one @graph with Organization, WebSite, Article, Person nodes |
| Semantic headings | Defines passage boundaries for chunking | One H1, logical H2–H4 nesting, no styling-only headings |
| Text-based facts | Numbers inside images are unreadable | Move data out of graphics into HTML tables |
| Explicit crawler policy | Controls which assistants may use your content | Deliberate robots.txt directives per AI user agent |
| Stable URLs and dates | Freshness and identity resolution | Visible publish and update dates in markup |
Entity strategy: be a thing, not a string
Assistants resolve queries to entities before retrieving documents. If your organisation is not resolvable as a distinct entity, you are competing as an anonymous text match. Three moves fix this: publish consistent Organization markup with sameAs links to your authoritative profiles; give every author a real page with credentials and external corroboration; and cover the concepts adjacent to your subject in an interlinked cluster rather than one long page, because entity coverage is measured across a site, not within a document.
Deciding your AI crawler policy
You have three defensible positions and one indefensible one. You can allow AI crawling broadly to maximise citation reach; allow retrieval-time crawlers while disallowing training crawlers; or restrict access and accept reduced presence. The indefensible position is having no policy, which usually means blocking the crawlers that would have cited you while leaving genuinely abusive scrapers untouched. Distinguish user agents used for live answering from those used for model training, and document the decision so marketing and legal stop relitigating it.
Measurement: new KPIs, not new dashboards
Traditional rank tracking cannot see an answer surface. Instrument these instead:
- Citation share. For a fixed set of your fifty most commercially important questions, ask each major assistant monthly and record whether you are cited, and alongside whom.
- Factual accuracy. Does the assistant describe your services, pricing model and positioning correctly? Misdescription is a content defect you can fix.
- Referral traffic from assistants. Segment it; volume is small, intent is unusually high.
- Branded query lift. Assistant exposure often shows up as branded search growth rather than direct clicks.
A 60-day implementation plan
Weeks 1–2: baseline. Ask 50 target questions across ChatGPT, Claude, Perplexity and Google AI features. Record who is cited and what is said about you. This document becomes your roadmap.
Weeks 3–4: fix retrieval. Confirm server-rendered content, publish the schema graph, resolve heading structure, set an explicit crawler policy.
Weeks 5–8: rewrite the twenty pages tied to those questions using the passage pattern, add author pages with credentials, and cite primary sources so your claims are checkable.
Then repeat the baseline. AEO is not a campaign with an end date; it is a monthly measurement habit attached to ordinary editorial discipline.
Frequently Asked Questions
What is answer engine optimisation?
AEO is the practice of making content retrievable, extractable and citable by AI assistants such as ChatGPT, Claude, Perplexity and Google AI features. It optimises for being quoted as a source rather than only for a blue-link ranking position.
Is AEO different from SEO?
It builds on SEO rather than replacing it. Crawlability, structured data, authority and clear information architecture still matter; what changes is that passages, not pages, are the retrieval unit, and citation rather than click-through is the outcome.
How do AI assistants choose which sources to cite?
Broadly, they retrieve candidate passages relevant to the resolved question, then favour sources that state the answer explicitly, are internally consistent, carry identifiable authorship, and agree with other reputable sources.
Do we lose traffic if an assistant answers instead of sending a click?
Some informational clicks disappear. What remains is higher-intent traffic plus brand-level influence, because being the cited source shapes the answer even when nobody clicks. Measure citation share alongside sessions.
How long does AEO take to show results?
Structural fixes such as extractable answer blocks and clean markup often surface within weeks; entity-level trust and consistent citation build over quarters.