How to get cited by Perplexity

Perplexity is the most source-forward of the big AI engines. It runs a web search for almost every question and shows the sources as numbered links, right next to the sentences they support. For a brand that cuts both ways. A citation there is visible and clickable. If you're missing, the reader can see exactly who got the spot instead.

The work overlaps with getting cited by ChatGPT, but the details differ enough that we run Perplexity as its own track. This is the order we go in.

Step 1: let PerplexityBot in, then check the logs

Perplexity documents its crawlers on its own site. PerplexityBot is the one that indexes pages so they can surface in answers. Perplexity-User fetches pages when a user's request calls for it. Perplexity also publishes IP ranges for its bots, which matters when you need to tell a real visit from someone spoofing the user agent.

Run three checks:

robots.txt doesn't disallow PerplexityBot, either by name or through a wildcard rule that covers the folders you care about.
Your CDN or firewall isn't challenging or blocking it. A zone-wide "block AI bots" setting often catches the search crawlers along with the training crawlers.
Your logs show PerplexityBot fetching your important pages and getting 200 responses.

The third check is the one people skip. A permissive robots.txt with a CDN rule quietly blocking the bot behind it looks fine from the outside. Only the logs tell you the truth.

Step 2: pick the prompts where Perplexity matters to you

Perplexity's interface is built for research. Sources come first, and suggested follow-up questions sit under every answer. So the prompts worth tracking there tend to be specific and a bit long: "which payroll tools handle contractors in both Germany and the UK", or "does [product] support SSO on the cheaper plans".

Here's a cheap trick for widening the set. Run each of your core prompts in Perplexity and read the related questions it suggests underneath. Those suggestions show you where a conversation goes next, and several of them will be prompts your buyers ask that never made it onto your keyword list. Add the relevant ones, drop the rest.

Step 3: read the source list like a competitor report

For every prompt where you're absent, open the sources and sort them. Your own pages. Competitor pages. Publishers and review sites. Forums and community threads. Video. You'll start to see a pattern per topic, and it changes the plan. A prompt cited mostly from vendor docs needs a better page from you. A prompt cited mostly from publisher roundups needs you on those roundups.

Then go one level deeper. Because Perplexity attaches citations to individual sentences, you can see which claim each source was used for. One source supplied the price range. Another supplied the definition. A third backed up "works well for small teams". This tells you what kind of statement gets lifted for that prompt, which is much more useful than knowing which domain won.

Ask a blunt question about each cited competitor page: is it better than ours, or just more specific? Most of the time it's the second. It answers the exact question in one clear paragraph, while your page answers a broader question across three screens.

Step 4: build pages around claims, not keywords

Citations attach to claims, so a page wins by containing the claim in a form that's easy to quote. What that looks like in practice:

One question per section. Put the answer in the first two sentences under the heading, then explain.
Specific facts you can stand behind. Supported countries, integrations, limits, the pricing model. If you don't publish prices, say how pricing works and who it suits. A vague page gives the engine nothing to attach a number to.
Visible dates. Perplexity doesn't publish how it weighs freshness, so we won't pretend to know. But a page that shows when it was last checked removes one obvious reason to cite someone else on a time-sensitive question.
Content in the HTML. If the key copy only appears after client-side JavaScript runs, a crawler may fetch a near-empty page.
Fair comparisons. A comparison page that admits where a competitor is stronger reads like a reference. One that never concedes anything reads like an ad, and we'd expect it to be a weaker source for a "which should I pick" question.

Step 5: get onto the domains Perplexity already trusts for your topic

After step 3 you'll have a list of domains that show up across many prompts. That list is your outreach plan. It might be two or three review platforms, a trade publication, an integration partner's directory, and a handful of community threads.

Treat each type differently. For review platforms, make sure your profile is complete and current. For publishers, pitch the specific roundup that keeps getting cited, with facts the writer can check. For community threads, answer honestly under your own name and follow the community's rules. A disclosed answer from someone at the company is fine in most places. Undisclosed promotion gets removed and remembered.

This part is slower than editing your own pages. It's also often where the competitors who keep beating you have their edge.

Step 6: track it on a schedule

Running prompts by hand once gives you a snapshot. Perplexity answers shift between runs and over weeks as new pages get indexed, so the same prompts need to run on a schedule before you can say anything changed.

We do this for clients in Promptwatch, and the feature that does the heavy lifting here is citation analytics. For each tracked prompt it records which pages and domains Perplexity cited, next to the same data for ChatGPT, Gemini, Claude, and the others, and citation trends show how those sources move over time. That turns step 3 from a one-off spreadsheet into something you can watch. Perplexity tracking needs a paid plan, since the free Explore tier covers ChatGPT only.

Agent Analytics closes the loop on step 1. It reads AI crawler logs through CDN integrations and shows whether PerplexityBot fetched a page, whether it hit errors, and whether that page later turned up as a citation. That's the split that saves the most time. A page that gets crawled and never cited has a content problem. A page that never gets crawled has an access problem, and no rewrite will fix it.

Once the first fixes ship, go back to step 3 with fresh data. The source list will have moved, and so will the next thing worth doing.