Perplexity answers from a live web search and shows exactly which pages it used, so if you're not in the Sources panel, you're not in the answer. It comes down to four checkable things: whether PerplexityBot can fetch your pages, whether they're indexed, whether they answer the question directly, and whether a stronger source beat you to the citation.
This is the diagnostic version of the broader playbook on getting recommended by ChatGPT: access, plain-HTML facts, a clear identity and a direct answer apply across every engine, but Perplexity is the one place you can watch the citation decision happen in the open, because it shows its sources.
Check this first, because nothing else on this page matters if the crawler can't fetch your pages at all. Perplexity documents PerplexityBot as the crawler that builds its index ahead of time, and states that it respects robots.txt. A Disallow rule under its user-agent, or a CDN's bot-protection setting that challenges AI crawlers by default, both block it silently, without any error a visitor would ever notice.
Proof: Perplexity — crawler documentation (PerplexityBot & Perplexity-User).
Check it now: open yoursite.com/robots.txt and look for a Disallow under PerplexityBot, then run the free AI Crawler Access Check, which also fetches your homepage with PerplexityBot's real user-agent to catch a CDN-level block robots.txt can't show.
No, and the difference matters if you've only checked one of them. PerplexityBot indexes pages in advance. Perplexity-User is a separate, real-time fetch triggered the moment someone asks about a specific page or pastes its URL into a chat, and Perplexity describes that one as acting on a user's request rather than as a bot, so it isn't bound by robots.txt the same way. Allowing PerplexityBot is still the right, controllable move. It is the difference between being indexed on purpose and only being read when someone happens to ask about you directly.
Good to know: see PerplexityBot's full crawler profile for its exact user-agent string and IP range, so you can confirm a request in your logs is really Perplexity and not an impostor.
A page PerplexityBot can fetch still has to be indexed before it's eligible to be cited, and a brand-new or rarely-linked page can sit outside that index for a while. Being crawlable is the entry ticket, not the citation itself. A page nobody links to, that sits deep in a site with no internal links pointing to it, or that only recently went live, is slower to show up as a candidate source at all.
| What's wrong | How to tell |
|---|---|
| PerplexityBot blocked | Disallow rule in robots.txt, or a CDN AI-bot block |
| Page not indexed yet | New or poorly-linked page, no internal links pointing to it |
| Page doesn't answer directly | The answer is buried past the first paragraph, or missing |
| Stronger source exists | A competitor's page states the answer more directly or more recently |
Do this: link to the page from somewhere that's already indexed, your homepage, a relevant blog post, or your sitemap, and give it a few days before concluding it's blocked rather than simply new.
Yes. Perplexity pulls whichever passage answers the question most directly, in plain HTML, not the page that's broadest or best-known. If the exact answer to "best X for [use case]" or "X vs Y" sits three paragraphs into a general page instead of stated up front on its own dedicated page, Perplexity has to work harder to find it, and usually cites a competitor's clearer paragraph instead. Content that only appears after JavaScript renders is invisible to PerplexityBot entirely, since it reads raw HTML.
Check it now: run the free What AI Actually Sees check on your key page to confirm the answer exists in the raw HTML PerplexityBot fetches, not only in a part of the page JavaScript builds afterward.
Then that's the real reason, and it's worth knowing rather than assuming you're blocked. Perplexity rebuilds its Sources panel from scratch for every question, so a narrower, more recently updated page with specific numbers and dates regularly out-cites a bigger, older "authority" page on the same topic. Independent reviews, comparison sites, and mentions elsewhere also feed into which claim Perplexity trusts enough to cite. The mechanics of how it weighs all of this are covered in full on how Perplexity chooses which sources to cite.
Bottom line: access gets you eligible. A direct, specific, recently-dated answer is what actually wins the citation once you are.
None of these require buying anything, and each takes a few minutes:
Bottom line: a missing citation is almost always one of four things, and all four are things you can check yourself today.
Usually one of four things: PerplexityBot is blocked from fetching your pages, your page isn't in Perplexity's index yet, your page doesn't answer the question as a self-contained passage, or a stronger, more direct source exists and gets cited instead.
Open your robots.txt and look for a Disallow rule under the PerplexityBot user-agent, then check your CDN or WAF's bot settings for an AI-bot block, since that layer doesn't show up in robots.txt. The free AI Crawler Access Check runs both in about a minute.
Yes. PerplexityBot builds its index ahead of time and documents that it respects robots.txt. Perplexity-User is a separate, real-time fetch made when someone asks about a specific page in a chat, and Perplexity describes it as acting on a user's request rather than as a bot.
Perplexity rebuilds its Sources panel from a live search on every question, so it picks whichever page states the answer most directly and reads as freshest at that moment. A page that buries the answer three paragraphs down loses to a competitor's page that states it in the first sentence, even if your page is otherwise more complete.
No. Access is the first requirement, not the only one. A page PerplexityBot can fetch still needs to be indexed, answer the specific question directly, and hold up against whatever else a live search turns up for that wording.
The free check reads your crawler access and the rest of the 23 checks — evidence included, nothing installed.