How Perplexity picks the sources it cites

A classic search engine returns links and lets you read. Perplexity runs a search, reads a small set of pages itself, and writes an answer that cites the specific pages it drew claims from. Those numbered citations are not decoration: they mark the sources whose sentences the answer is built on.

That changes what you are optimising for. Position on a results page is not the target, because Perplexity is not showing a results page. Being one of the handful of sources it retrieves and then finding your sentence clear enough to quote is the target. The visible citations make this unusually measurable, since you can read any answer and see whether you were named.

Perplexity documents two user-agents. PerplexityBot builds and maintains the search index over time, and Perplexity-User fetches pages live on behalf of a person during an active question (Perplexity, 2026). The distinction matters for the same reason it does with other engines: one keeps a standing index of you, the other reads you in the moment a user asks.

Access is governed by robots.txt, which the Robots Exclusion Protocol defines and compliant crawlers are expected to honour (IETF, 2022). Perplexity states that if you disallow PerplexityBot, it will not index the text of your pages, though it may still show your domain, your headline, and a brief factual summary (Perplexity, 2026). So a block is not a clean disappearance: it removes your quotable text while leaving a thin reference behind.

The practical first step is to confirm you are actually retrievable. Your important pages need to be reachable, allowed in robots.txt for the Perplexity agents you want, and served as real content in the HTML rather than assembled only after JavaScript runs. A page a crawler cannot read is a page that cannot be cited.

Three signals show up again and again in which pages get cited. The first is relevance to the specific sub-question, not the broad topic: Perplexity is filling a particular slot in its answer, and it wants the page that addresses that slot directly. The second is recency, which matters most for questions whose answers change over time, where a current page beats a stale one making the same claim. The third is source authority, where a claim carried by a source Perplexity treats as reputable is a safer thing for it to repeat.

It is worth being honest about which of these is documented and which is observed. Perplexity documents its crawlers and its robots.txt behaviour. The weighting of relevance, recency, and authority is what practitioners infer from watching which pages get cited across many answers, and it is consistent enough to plan around, but treat it as informed observation rather than a published ranking formula (Osoro Solutions, 2026).

Start with retrieval, because nothing else matters if you are not in the set. Confirm indexing and crawler access as above.

Then make the page answerable. Because citations are assigned while the answer is written, Perplexity needs a clean, self-contained sentence it can lift for the question at hand. Open each section with the direct answer to its heading, phrase the heading the way a person asks the question, and keep the claim specific enough to be worth quoting. A page that buries its answer in a long build-up gives Perplexity nothing to cite even when it retrieved the page.

Then keep it current. For any page answering a question whose answer moves, a genuine update with a visible date is a real signal, not a cosmetic one. Do not fake it, but do maintain the pages that matter.

Finally, build corroboration. Authority here is partly about being referenced and discussed on the sources Perplexity's retrieval respects, which is the slow off-page work of earning mentions rather than buying links. It is the same discipline that helps you across every answer engine, and it compounds.

This is a genuine decision, not an obvious one. If your priority is keeping your content out of Perplexity's index, blocking PerplexityBot does that for the text. But remember the documented behaviour: Perplexity may still surface your domain, headline, and a short summary even when blocked, so you do not vanish, you just stop being quotable (Perplexity, 2026).

For most brands that want to be recommended inside answers, blocking is a cost rather than a protection, because it removes exactly the quotable material that earns a citation. Decide it deliberately, set robots.txt to match, and do not let a blanket AI-crawler block make this choice for you by accident.

Perplexity makes this easier than any other engine. Ask your real buyer questions, read the answer, and note which domains are cited and for which part of the answer. Do it on a schedule so you can see movement, and record when a competitor holds a slot you want. There is no analytics event for a citation, so this direct sampling is the measurement, and with Perplexity the evidence is right there in the numbered links.

Work in dependency order. First the retrieval checks, because they gate everything. Then the editorial pass on the pages that matter most to your pipeline, opening each section with its answer and phrasing headings as questions. Then the ongoing work of freshness and corroboration, which is slower but is what separates a page that gets cited once from a source Perplexity returns to. Measure by reading the citations, and let what you see there tell you which question to win next.