How to Rank in Perplexity: The Citation Playbook

Of the big AI answer engines, Perplexity is the one that shows its working. Ask it which tool a buyer in your category should use and it does not merely name a shortlist, it prints the sources it read to build that shortlist, so where ChatGPT and Gemini keep their citations behind the curtain, Perplexity hands you the reading list. That single habit changes the whole exercise, because ranking in Perplexity is less a matter of guessing what the model likes and more a matter of reading which pages it already trusts for your queries and then getting your brand into that short list. The case of this piece is straightforward: you do not rank in Perplexity by polishing your homepage, you rank by earning your way into the handful of community and third-party sources it cites, and then by using its own transparency to check whether the work actually landed.

None of that is search engine optimisation in the old sense, and it helps to name the thing plainly: this is AEO (answer engine optimisation, or getting the robot to name you before it names anyone else), and Perplexity is the friendliest surface on which to practise it, precisely because it tells you when you are failing. The figures below come from our own measurement rather than a vendor deck, and they explain why the tactics that move Perplexity are not the ones that move an old blue-link ranking.

Perplexity is the community surface

On 13 July 2026 we at Honeyb (this is our product) ran 20 buyer-intent prompts three times each across four engines through their APIs, which produced 240 answers and 2,507 citations to pick apart, and Perplexity's Sonar model behaved unlike the rest. Of the 498 sources it cited, 71 came from Reddit and 40 from YouTube, so a little under a quarter of everything it leaned on was community content rather than polished marketing pages. Profound's larger analysis finds the same shape from a different angle, putting Perplexity's share of citations drawn from social and community sources at 19.4%, against low single digits for ChatGPT, Claude and Gemini. The lesson for a marketing team is uncomfortable but useful, which is that the pages getting you named in Perplexity are mostly ones you neither own nor can simply publish.

So roughly one Perplexity citation in five is a place you can influence only by being genuinely useful in public, which is why the playbook that follows starts with earning mentions and treats your own pages as the last mile rather than the first.

Read its citations; that is the playbook

Here is the move that makes Perplexity worth singling out from ChatGPT and Gemini, both of which tend to hide the sources behind their answers. Perplexity lists them, on every answer, which means you can run the exact prompts your buyers use and read back the precise pages, threads and videos it consulted to reach its shortlist, so a guessing game becomes an audit: instead of theorising about what the model wants, you collect the twenty or thirty sources it actually cites across your category's questions, sort them by how often they recur, and you are looking at your real target list.

Do that and two things usually fall out. The first is a small set of high-recurrence sources, a particular subreddit, one or two comparison articles, a directory, a YouTube channel, that show up again and again across related prompts, and those are where a single mention earns the most leverage. The second is the quiet discovery that your own pages are being read and their facts reused while the credit goes elsewhere, which is close to the national sport of AI search, since 62% of AI citations never name the brand being cited, so your words can be doing the work while a review site collects the recommendation. Knowing that shifts the job from writing more pages to getting your name attached to the claims those pages already make, and if you want the outreach mechanics in detail our guide on how to get cited by AI walks through them; the point here is that Perplexity hands you the target list for free.

Want to get recommended by AI?

Check your AI search visibility, then let the Honeyb agent write, fix, and earn what gets you recommended. Free to start.

Free AI visibility checker

Earn the mentions, do not manufacture them

Once the audit tells you which sources Perplexity trusts, the work splits three ways, and the order matters because the leverage is wildly uneven. Community mentions come first, because Reddit is both Perplexity's largest single community source and, on Semrush's numbers, the most-cited source across all of AI at 40.1% of citations, and you cannot buy your way into a thread, so the only durable tactic is to answer real questions in your category's subreddits under a disclosed affiliation and treat a genuinely useful reply as a content asset rather than an advert, because the moment it reads as marketing it is removed, and rightly. Video comes second, since a plain, specific walkthrough or an honest head-to-head becomes a citable YouTube source, and Ahrefs is blunt that AI visibility correlates most strongly with third-party mentions and video rather than with anything you do on your own page. Third-party reviews and directories come third, and they are the most straightforward of the three, because getting listed and fairly described in the comparison posts and tool directories that already rank is slow but entirely within reach.

If you have the budget for exactly one of the three this quarter, spend it on the community and video work, because that is where the data says the movement is, and because a directory listing you can chase at any time.

Your own pages still matter, just less than you think

None of this makes the website irrelevant, because roughly 78% of Perplexity's citations were still ordinary web pages, and because the model retrieves live at the moment of the question rather than from a months-old index, which rewards two things in particular. Freshness is the first, since a page dated this quarter with its claims current is easier for a live-retrieval engine to trust than one that could be describing last year's product. Structure is the second, because engines lift tables and direct answers with unreasonable enthusiasm, so a page that opens with the answer and then gives a clean comparison table is far more liftable than the same facts buried in three paragraphs of preamble. The verdict is one of proportion: on-page work is the smaller lever, worth doing well but not worth mistaking for the whole job, and a team that spends a month perfecting its homepage while ignoring the subreddit Perplexity keeps citing has polished the one surface that moves the answer least. For the equivalent tactics on the other two big surfaces, our guide to ranking in ChatGPT, AI Overviews and Perplexity sets them out one engine at a time, and how Perplexity AI works explains the retrieval mechanics underneath all of it.

Measure it, because one check will lie to you

The last discipline is the one most teams skip, and it is the one that separates a real gain from wishful thinking, because AI answers are not stable and a single glance at a single Perplexity result tells you almost nothing. In our test Perplexity changed its top recommendation between two identical runs 43% of the time, and the brand list it produced overlapped only 61% from one run to the next, while SparkToro's wider study finds the same question returning a materially different answer roughly 70% of the time. Read that as a warning about method, because check once and you might catch a flattering answer that will not repeat, or a dismal one that was a fluke. The honest metric is recommend rate, the share of repeated runs in which your brand actually appears, measured before you start and again after each change, which is why spot-checking fails as a way to track this and why a tool that runs your prompts on a schedule earns its keep; our roundup of Perplexity monitoring tools compares the ones that trace citations rather than merely handing you a score.

The shortest path to a baseline is to see what Perplexity and the other engines say about you right now, before you spend a week writing subreddit answers. Run a free AI visibility check to find out whether you are being named, merely cited, or missed altogether, and which sources are writing the answer in your place.