How to rank in AI search: optimising for LLM answer engines.

Answer engines retrieve passages, not pages

Where classic search ranks a URL as a whole, retrieval-augmented answer engines pull specific passages that directly answer the query, often just a paragraph or two. A page can rank poorly overall yet still get cited if it contains one exceptionally clear, self-contained paragraph that directly answers a common question.

This changes the writing unit from the page to the paragraph. Every section of a page should be able to stand alone as a complete answer to the question its heading implies, without depending on context from paragraphs above it.

Directness and specificity beat persuasive framing

Answer engines favour content that states a clear position or number early and explains the reasoning after, rather than content that builds up to a conclusion through narrative framing. A paragraph that opens with a hedge or a rhetorical question is less likely to be extracted cleanly than one that opens with the actual answer.

Specificity beats generality in the same way it does for human readers, but the effect is sharper for extraction: a paragraph naming an actual number, timeframe or named entity is easier for a model to lift as a confident answer than one full of qualifiers.

Source credibility signals still gate citation

Answer engines weight the credibility of a source before extracting from it, using signals that overlap heavily with traditional authority signals, consistent publishing history, clear authorship, citations from other credible sources. A technically perfect paragraph on a site with no established credibility is less likely to be surfaced than a good-enough paragraph from a recognised authority in the space.

This means the entity-level schema and author transparency work described elsewhere is not a side project, it is part of the same visibility system as the content itself.

Structure content around the actual questions people ask an assistant

The phrasing people type into a search box differs from the phrasing they speak or type to a conversational AI assistant, which tends to be longer and more natural. Research the conversational variants of a topic's core questions, not just the keyword phrasing, and address them explicitly with their own heading and self-contained answer.

There is no dashboard yet, so build a manual tracking habit

Measurement tooling for AI citation is immature compared to search rank tracking. In the meantime, run a rotating set of representative queries through the major AI answer engines monthly and log whether and how the business is cited, including whether the citation is accurate. This manual process is unglamorous but currently the most reliable way to know if the work is moving the needle.