How does Perplexity decide who to cite?

Search first, then synthesis

Perplexity describes itself as an answer engine, and mechanically it runs the sequence this whole series keeps returning to: the question triggers live retrieval from a web index, a set of top results is read, and the answer is synthesised with each claim tied to a numbered source. Perplexity documents two agents: PerplexityBot, which builds the index behind cited answers, and Perplexity-User, which fetches a page when a user's question sends the engine to it directly. Per Perplexity's documentation, robots.txt is the control surface for how your site interacts with them, so the routine first check applies here too: know what your rules currently say, and note that shutting PerplexityBot out removes you from Perplexity's cited answers entirely, which some publishers do deliberately and no visibility-seeking SME should do accidentally.

Two structural facts follow from the retrieval-first design. First, ranking somewhere for the topic is the entry condition; Perplexity synthesises from what retrieval surfaces, and a page retrieval never sees cannot be cited. Second, because every sentence of the answer wants a supporting reference, the competition is not for a position on a page but for the role of best support for a specific claim. That reframing is genuinely useful when you plan content: you are not writing pages that rank so much as writing claims that deserve citing.

What the numbered citation rewards

Checking Perplexity's answers by hand across client categories, the cited pages share four properties, each of them buildable.

Freshness that is visible. Perplexity leans noticeably toward recent material, and toward pages whose recency is machine-legible: honest dates in the content and the metadata, updated-on signals, and substance that reflects the current state of the topic. This is the strongest argument in the whole series for the refresh-not-replace discipline this very page runs on: a guide that is genuinely re-reviewed, on a stable address, with its date declared, is built the way Perplexity likes its sources.

Claims with support inside them. Because the engine attaches references to claims, it favours pages where the claim and its evidence sit together: the number with its context, the recommendation with its reason, the definition with its authority, the source named and linked. Pages that assert without support give the synthesiser nothing safe to lean on.

Direct answer structure. The familiar discipline, at its most literal: question-led headings, complete answers in the opening sentences, liftable without surgery. The overlap with our AEO work is total, because answer-engine formatting is exactly what is being rewarded.

Recognised authority for the claim type. Perplexity runs a Publishers' Program with named media partners for news-flavoured queries, which an SME neither can nor needs to join. For commercial and category questions, the authority test is the one the rest of your search visibility already runs on: does the wider web corroborate that this source knows this subject? Earned authority transfers; it is the same asset, cited by one more engine.

The standing caveats: answers vary between runs, so presence is a probability you improve rather than a position you hold; and Perplexity ships product changes quickly, so this page describes the engine as at its last-reviewed date and gets updated here, not replaced.

Why this engine punches above its size

Perplexity's user base is smaller than Google's by orders of magnitude, but it is dense with exactly the person a proof-led business wants: the researcher who checks sources before deciding. When that person's answer cites you, they are one click from your evidence with their scepticism already half-spent, because the engine vouched for the reference. For a business whose whole position is verifiable proof, and ours is, that is a disproportionately valuable surface per user reached.

What we do about it, and what you can do today

The Perplexity work is the AI-search work with the dials set to citation: fresh, dated, genuinely maintained pages; claims written with their evidence attached; answer-shaped structure; authority earned where the engines look for it; and manual answer checks, because watching what the engine actually cites for your buyers' questions is the only measurement that matters. The AI-search visibility check includes Perplexity, free, human-reviewed.