Perplexity is the answer engine that shows its work. Every answer comes with numbered citations, a sources panel, and a follow-up box — which makes it the most measurable AI surface you can optimize for. If you want to see whether AI visibility work actually works, Perplexity is where you look first.
It is also the surface most sites accidentally opt out of. Perplexity’s crawler setup has a quirk that removes careless sites from its index while their owners believe they only blocked “AI training.” This playbook covers how Perplexity sources answers, what its crawlers actually do, what citation-worthy content looks like here, and how to test whether any of it is working.
Key Takeaways
- Perplexity is retrieval-first: every answer is built from live web sources with numbered citations — there is no training-data-only mode to worry about.
- PerplexityBot is the search crawler that decides whether you can be found; Perplexity-User is the user-triggered fetcher that generally ignores robots.txt. Know which one you are configuring.
- Perplexity states PerplexityBot is not used to crawl content for AI foundation models — blocking it is a visibility decision, not a training opt-out.
- Citation-worthy content here looks like everywhere else: extractable definitions, original data, comparison tables, and visible freshness — plus a title that survives being read as a source label.
- Nobody can guarantee a Perplexity citation. The playbook maximizes eligibility and extractability; the rest is the system’s judgment.
How Does Perplexity Actually Source an Answer?
Perplexity’s model is simpler than ChatGPT’s — and that simplicity is good news for optimizers.
There is no training-data mode to untangle. When you ask Perplexity a question, it searches the live web, retrieves relevant pages, and synthesizes an answer with numbered citations pointing at the sources it used. Every claim in the answer is, in principle, traceable to a source. The sources panel lists them; follow-up questions keep the same source set in context.
That retrieval-first design has two consequences for your strategy. First, your page must be in Perplexity’s index — there is no path to being cited from training data. Second, the citation is the product, not an afterthought: Perplexity’s whole interface is built around showing where each part of the answer came from, so citation-worthy formatting is citation-generating formatting.
The practical upshot: if you are invisible on Perplexity, the cause is almost always technical (crawler blocked, page not indexed) or structural (content not extractable) — not mysterious. Both are fixable, and this playbook fixes both.
PerplexityBot vs Perplexity-User — What’s the Difference?
Perplexity documents two user agents, and they do completely different jobs. Confusing them is the most common technical mistake on this platform.
| PerplexityBot | Perplexity-User | |
|---|---|---|
| Job | Crawls and indexes the web so sites can surface and be linked in Perplexity search results | Fetches a specific page when a user’s question needs it |
| Used for AI training? | No — Perplexity states it is not used to crawl content for foundation models | No — not used for crawling or training |
| robots.txt | Follows robots.txt; Perplexity recommends allowing it | Generally ignores robots.txt, since a user requested the fetch |
| Effect of blocking | Your site is not indexed for Perplexity search — no citations, ever | Only blockable at the firewall; stops Perplexity quoting you accurately in those conversations |
The practical reading: PerplexityBot is the one that decides whether you can be found. If you blocked it while trying to block “AI crawlers” generally — the mistake my robots.txt guide exists to prevent — you removed yourself from every Perplexity answer while believing you only opted out of training. There is no training data to opt out of here; blocking PerplexityBot is purely a visibility loss.
Perplexity-User is the subtler case. Because it acts on behalf of a specific user asking a specific question, it generally ignores robots.txt — much like a browser fetching a page for a person. Blocking it at the firewall is possible, but it stops Perplexity from quoting your page accurately when users ask about it directly. For most businesses that want visibility, allow both.
Perplexity publishes its crawler IP ranges (perplexitybot.json and perplexity-user.json), so you can verify that requests claiming to be Perplexity actually are — scrapers borrow well-known user-agent strings to inherit their welcome. And crawler-setting changes take roughly 24 hours to propagate, so do not test five minutes after editing robots.txt.
What Does Citation-Worthy Content Look Like on Perplexity?
Perplexity cites like a researcher with a deadline: it wants sources that make each sentence of its answer defensible. The shape of citation-worthy content will be familiar if you have read my ChatGPT citations playbook — the mechanics differ, but the content shape is the same discipline:
Lead with extractable definitions. Perplexity’s answers often open with “what is X,” and the clearest definition wins the first citation. One crisp, quotable definition in your opening paragraph — standalone sentences a system can lift verbatim.
Publish original data with methodology. The most-cited pages in any niche are the ones with numbers nobody else has. Methodology matters as much as the number: sample sizes, dates, and how the data was collected. A naked percentage is not citable; a documented finding is.
Use tables for comparisons. “X vs Y” tables compress exactly the structure an AI answer needs. Prose comparisons almost never get cited as cleanly — rebuild them as tables and keep the prose for depth.
Write titles that survive as source labels. On Perplexity, your page title appears next to its citation number in the sources panel. A title like “Home” or “Blog” tells the user nothing; a title like “AI Keyword Research Workflow: From Seed Keyword to Content Plan” tells them why to click. Titles are citation UI here — write them accordingly.
Show your freshness. Visible publication and update dates. For time-sensitive questions, Perplexity favors current sources, and a maintained page beats an identical abandoned one.
Structure in Q&A. Question-form headings with direct answers map directly onto how Perplexity assembles its responses. Your H2s are pre-written answer components.
None of this guarantees a citation. It makes your page the kind of source a retrieval-first system is designed to prefer — which, on Perplexity more than anywhere else, is the entire game.
How Should robots.txt Be Set Up for Perplexity?
The configuration is simple — the decision is what matters:
If you want Perplexity visibility (recommended for most sites):
# Perplexity search crawler: allow
User-agent: PerplexityBot
Allow: /That is it for the search side. Perplexity-User cannot be meaningfully controlled via robots.txt (it generally ignores it), so no rule is needed — and adding a Disallow for it gives a false sense of control. If you truly need to block user-triggered fetches, that is a firewall decision, not a robots.txt one.
If you want out of Perplexity entirely:
User-agent: PerplexityBot
Disallow: /Be honest about what this buys you: no citations, no source-panel presence, no referral traffic from Perplexity answers. Since PerplexityBot is not a training crawler, this is not a training opt-out — it is purely a visibility sacrifice. Make it deliberately or not at all.
After changing: allow roughly 24 hours for the change to propagate before testing, and verify with your server logs — look for the PerplexityBot user-agent string (and cross-check the IP against Perplexity’s published ranges) rather than trusting the file alone.
How Do You Test Whether Perplexity Cites You?
Perplexity is the easiest AI surface to test on — its citations are right there in the answer. Use that:
Run a prompt portfolio. Pick 15–30 questions your buyers actually ask. Run them monthly in a clean session, logged out. Log: did an answer appear, were you cited, which page was cited, and which competitors appear. The numbered citations make this faster than on any other platform — you can see your position in seconds.
Vary the question phrasing. Perplexity retrieves per query, so rephrased questions surface different source sets. If you are cited for “how to track AI citations” but not “AI citation tracking tools,” your content matches one phrasing and misses the other — a coverage problem, not an authority problem.
Check the sources panel, not just the answer. The inline numbers show what made the final cut; the sources panel shows the full retrieved set. If you appear in the panel but not inline, your content was considered but not chosen — usually a sign the extractable unit (definition, table, data point) needs sharpening.
Watch for the residue in analytics. Perplexity citations that earn clicks arrive as referral traffic from perplexity.ai. The same GA4 AI-traffic discipline applies — and as elsewhere, clicks undercount citations, so treat referral sessions as a floor.
Consider a tracker at scale. When the prompt set outgrows manual testing, the brand-mention tracking methods cover the tools that automate citation extraction across engines, Perplexity included. Run any tool alongside your spreadsheet for a month before trusting its numbers.
Final Thoughts
Perplexity is the honest broker of AI search: it shows its sources, numbers them, and links them. That transparency makes it the best place to learn whether your AI-visibility work is real — if you cannot get cited on the platform that shows its work, the problem is in your content or your crawler setup, not in the algorithm’s opacity. Check your robots.txt today (ten minutes), restructure one page for extraction this week, and run your first prompt portfolio this month. For the full framework this playbook sits inside, read the brand visibility pillar and its sibling, the ChatGPT citations playbook.
People Also Ask
Does Perplexity use my content to train AI models?
Per Perplexity’s own documentation, PerplexityBot is not used to crawl content for AI foundation models, and Perplexity-User is not used for crawling or training either. Blocking PerplexityBot is a visibility decision, not a training opt-out.
Can I block Perplexity-User with robots.txt?
Not reliably. Perplexity states that Perplexity-User generally ignores robots.txt because the fetch was requested by a user. Blocking it requires firewall-level rules — and doing so stops Perplexity from quoting your pages accurately in user conversations.
How long until Perplexity reflects my robots.txt changes?
Roughly 24 hours for crawler-setting changes to propagate, per Perplexity’s documentation. Verify in your server logs rather than re-testing immediately.
Why does Perplexity cite my competitor but not me?
Usually one of three reasons, in this order: PerplexityBot is blocked or your page is not indexed; your content does not match the query’s phrasing; or your page is less extractable than the competitor’s (weaker structure, no quotable facts, no tables). Diagnose in that order.
Do I need to rank on Google to be cited by Perplexity?
No — Perplexity runs its own retrieval over its own index. The content qualities overlap (structure, authority, freshness), but Perplexity citations and Google rankings are separate systems. A page invisible on Google can still be cited on Perplexity if PerplexityBot can reach it.
