How to get cited by Perplexity: two bots, one honest playbook

Perplexity runs two agents, and they follow different rules

Most guides talk about PerplexityBot as if it were the whole story. It is not. Perplexity's crawler documentation names two user agents. PerplexityBot is the indexer, described there as designed to surface and link websites in Perplexity search results, and explicitly not used to crawl content for foundation models. Perplexity-User is the agent that fetches a page in the moment somebody asks a question that needs it.

The difference is not cosmetic. The same documentation says PerplexityBot respects robots.txt and recommends allowing it, and says Perplexity-User generally ignores robots.txt because the fetch was initiated by a person rather than by a crawl schedule. If you block by user agent, you are only governing one of the two.

Where Perplexity's own two answers do not line up

Perplexity's help centre article on robots.txt takes a narrower line than its developer documentation. It says the feature that let users request a summary of a URL regardless of robots.txt has been disabled to prevent misuse, and that PerplexityBot now only crawls in compliance with robots.txt. Both pages are published by Perplexity. They do not fully agree about what a user-initiated fetch is allowed to do.

We are not going to pretend to resolve that, because we cannot. What we can say is that the practical move is identical under either reading: allow PerplexityBot deliberately, and never treat your robots.txt file as evidence of what actually reached your server. Both of those are covered in more depth in the AI crawlers guide.

Verify access below the robots layer

Perplexity publishes IP ranges for both agents, at perplexity.com/perplexitybot.json and perplexity.com/perplexity-user.json, so a firewall or bot-management product can allow them by address rather than by a user-agent string that anyone can spoof. This is the check that catches the failure nobody expects: a robots.txt that reads as perfectly welcoming, sitting above a security product that has been quietly challenging the crawler for months.

Run it in this order. Confirm there is no Disallow covering PerplexityBot. Confirm your firewall is not challenging the published ranges. Then confirm the agents appear as successful fetches in your server or CDN logs. Only when all three pass is access ruled out as the reason you are missing.

The unit you compete for is a numbered citation

Perplexity attaches numbered sources to claims inside the answer, not to the answer as a whole. That changes what a winning page looks like. You are not trying to be broadly relevant to a topic. You are trying to be the page that states one specific, checkable thing more plainly than anyone else states it, so that the sentence needing support points at you.

In practice that means the same discipline that works on ChatGPT, applied harder. Definitions in the first sentence. Numbers with dates attached. Prices published rather than hidden behind a form, because a page that names a figure is quotable and a page that says it depends is not. Comparisons in tables. If a machine has to infer your claim from three paragraphs of context, a competitor who wrote it in one line gets the number.

What nobody can tell you

Perplexity publishes crawler documentation, IP ranges and an API. It does not publish how sources are ranked for a given answer. Every circulating list of Perplexity ranking factors is inference, and most of it is inference from a handful of prompts run once. Treat any guide quoting a citation-rate benchmark with suspicion, including ours if we ever start: no credible public dataset exists for how often a given site gets cited, which is precisely why the honest method is to measure your own site rather than trust an industry average.

That is also the difference between AEO and the older acronym pile. If you want the vocabulary settled before going further, GEO, AEO and SEO explains what actually changes about the work.

Perplexity is the engine you can actually measure

Citations in Perplexity are clickable, so a citation that earns a click lands in your analytics as a referral from perplexity.ai. Compared with engines where a mention leaves no trace, that is unusually honest feedback. Two things follow. First, you can prove the work paid off without buying anything. Second, referral counts are a floor rather than the full picture, because being read and not clicked still shapes what a buyer believes before they ever reach you. Tracking both, cited versus clicked, is what AI Citation Monitoring runs monthly.

Citations are starting to carry money

On 26 August 2025 Perplexity announced Comet Plus, a five dollar per month subscription that shares revenue with participating publishers across three interaction types: human visits, search citations and agent actions, as reported by Search Engine Journal on the day of the announcement. The share retained for compute was not disclosed, and the programme is for partners rather than open to any site, so this is not a revenue line you can plan around today. It matters for a different reason. An engine that pays for citations has told you what it values, and the thing it named second was being cited.

A two-week version of this

Week one is access and evidence. Read your robots.txt for PerplexityBot, check your firewall against the published IP ranges, pull the last thirty days of logs for both agents, and open your analytics to see whether perplexity.ai already appears as a referrer. Most sites learn something uncomfortable in this step, and it costs nothing but attention.

Week two is one page. Pick the single buyer question you deserve to win, ask Perplexity that question, and read who it cites and what sentence each citation supports. Then rewrite your page so the equivalent sentence on your side is shorter, more specific and carries a number. Re-ask in seven days. That loop, one question at a time, is the whole practice, and it beats a rewrite of forty pages done on a hunch.

Common questions