How to Rank in Perplexity: Crawlers, Content and Citations

Perplexity is the most transparent of the AI answer engines about where its answers come from. Every response shows its sources, and people click them. That makes it the easiest engine to diagnose: you can see who is being cited instead of you, and usually work out why. Most of the time the reason is not clever. It is a crawler being blocked, or a page that does not quite answer the question.

The broad mechanics are the same as in our guide to ranking in ChatGPT, with one emphasis that matters. Perplexity is search-first. For a typical question it runs a live search, retrieves a set of pages, reads the relevant passages and writes an answer that cites them as numbered sources. The model's training data matters much less than which pages the search step finds.

That has three consequences. You need to be in the index Perplexity searches, which mostly means letting its crawler in. You need to be retrievable for the way the question is phrased, which is ordinary relevance. And once retrieved, your page has to contain a passage that answers the question cleanly enough to be quoted, because a page that is relevant but vague gets read and then not cited.

Perplexity documents two user agents, and confusing them is the most common reason sites make the wrong robots.txt decision. Our glossary entry on PerplexityBot has the short version.

PerplexityBot is the crawler for Perplexity's search index. Perplexity says it is used to surface and link websites in its results, that it is not used to crawl content for training AI foundation models, and that it respects robots.txt. This is the one that decides whether you are in the index most citations come from.

Perplexity-User fetches a page when a person's request needs it, for example when someone pastes your URL and asks a question about it. Because a person asked for the fetch, Perplexity says this agent generally ignores robots.txt, in the same way a browser would. It does not build the index.

Perplexity also publishes the IP ranges its crawlers use, which matters if you run a firewall that verifies bots by IP. Spoofed user agents are common, and a rule that allows anything claiming to be PerplexityBot is weaker than one that checks the published ranges.

Start here, because content work is wasted if the crawler cannot get in.

A minimal, permissive rule for PerplexityBot looks like this:

User-agent: PerplexityBot
Allow: /
Disallow: /account/
Disallow: /checkout/

Sitemap: https://example.com/sitemap.xml

Our free Domain Score Checker reads your robots.txt and reports which AI crawlers are allowed, which is a quick way to catch a rule you did not know was there. The wider picture of who crawls for what is in our entry on AI crawlers.

Once access is right, the content principles are the same ones that earn citations in Google AI Overviews, with a slightly stronger weight on freshness and specificity.

Answer the question in the first lines. Perplexity quotes passages. A heading that matches the question, followed by a direct two-to-four sentence answer, is the shape most likely to be lifted. Put explanation and nuance after it.

Be visibly current. Perplexity is used heavily for questions where recency matters: prices, tools, comparisons, rules. Show a real updated date, change it when the content changes, and keep time-sensitive facts accurate. A page that says "in 2024" in its first paragraph is easy to pass over for one that is up to date.

Say something checkable. Perplexity's users click sources to verify claims, so pages that state specific, attributable facts tend to be useful sources. Name the source of a number, link to the primary document, and avoid vague claims that a model cannot pin to anything.

Cover comparisons and alternatives honestly. Many commercial questions in answer engines are comparisons: which tool, which provider, X or Y. If you sell something, a fair comparison page that admits where a competitor is the better choice is more citable than a page that only praises you, because it is more useful to the person asking.

Do not gate the answer. If the useful part of a page sits behind a login, an email form or a cookie wall that hides the content until someone clicks, a crawler may see only the wall. Gated assets can still exist; just make sure the page that describes them says something substantive in the open, so there is a passage to cite.

Give pages honest titles and URLs. A title that says what the page answers, and a readable URL, help the search step match your page to the question before the model reads a word of it. A page called "Resources" with a URL full of IDs starts at a disadvantage against one called "How to choose a payroll provider for a five-person company".

Keep structure clean. Headings, short paragraphs, tables for genuinely tabular data, and accurate schema markup for dates and authorship. None of this is a Perplexity ranking factor that Perplexity has published. It just makes the right passage easy to find and quote. An llms.txt file is a reasonable low-cost addition on the same logic, though there is no public evidence that Perplexity uses one.

Perplexity is the easiest engine to check by hand, because the sources are on screen. Keep a list of the questions that matter to you, ask each one on a regular schedule, and record which domains are cited and whether you are among them. When a competitor is cited and you are not, open their page next to yours and compare the passage that was used. The difference is usually obvious: they answered the question in a sentence, and you answered it in a section.

In analytics, Perplexity clicks generally arrive as referrals from perplexity.ai, so they are visible in a way that some other AI traffic is not. Group them with the other answer engines so you can see the trend in one place; our guide to AI visibility tracking covers how to combine manual checks, referral data and tracking tools.

A word on expectations. Being cited in Perplexity is valuable and, for many topics, achievable with ordinary good practice. It is not a substitute for a business people already search for by name, and for some questions the answer is complete enough that few people click. Treat it as one channel among several, and measure the visits it brings by what they do next.