How to Get Cited by Perplexity AI

That split matters at Perplexity's scale. The company says people ask it 1.5 billion questions a month, and every answer shows numbered sources you can click. Those sources are concentrated. In Ahrefs' September 2026 snapshot of more than 3.1 million US queries, Reddit, YouTube and Wikipedia held 48.7% of the citations that went to Perplexity's 50 most-cited sites.

This guide explains what it takes to be cited by Perplexity AI. It starts with how Perplexity names and cites brands, then walks through the retrieval layer that picks a source for each prompt. You'll get what Perplexity documents, what studies have observed, a ladder for finding where your page drops out, and a reachability check. For the full technical playbook, read our Perplexity SEO guide.

How Perplexity Mentions and Cites Brands

Every Perplexity answer has two layers. The text names things, including brands. The citations link to the pages the text was built from. Perplexity's guide tells users to "highlight answer text to check the sources referenced," so each sentence can be traced to its sources.

Those two layers don't always line up. A "best invoicing tools" list may be the cited source, while the brands it lists are the ones named in the answer. Your own help article may be cited for a general fact, with no brand named at all. Our guide to brand mentions in Perplexity maps all four combinations and what causes each.

Why lists drive so many brand names

In an August 2026 study of 460 B2B prompts, the tracking vendor Analyze AI classified 115,843 citations by page type. For Perplexity, 35.4% went to brand websites and product pages, 34.4% to lists, comparisons and reviews, and 25.2% to editorial pages. On matched prompts, Perplexity named the brand being tracked in 41.7% of answers, against 32.3% for ChatGPT. The authors' explanation: one cited "top 10" article puts ten brand names in front of the model.

So there are two routes to being named. Your own page gets cited, or a page Perplexity trusts mentions you. A plan to get cited by Perplexity AI needs both.

Mentions hold steadier than sources

The 2026 paper "Don't Measure Once" ran the same prompts repeatedly within 24 hours. Perplexity's source lists overlapped by a Jaccard score of 0.282, where 1 means identical. The sets of brands it named overlapped by 0.492. Sources rotate faster than names. That's why you track both, and why our guide to tracking brand mentions in Perplexity logs each one separately.

What Perplexity Documents About Searching and Citing

Perplexity publishes more about its search stack than most AI vendors. Here is what its own pages say, as of October 2026.

The answer engine

Its help center says Perplexity "searches the internet in real-time," gathers information "from authoritative sources like articles, websites, and journals," and adds "numbered citations linking to the original sources." Pro Search runs multiple searches and shows "how the AI broke down your question." The Research mode performs "dozens of searches" and reads "hundreds of sources."

The breakdown is useful to you. Each sub-question Perplexity shows is a separate retrieval, and each one is a chance for a different page to win.

The index and the ranking pipeline

Perplexity's September 2025 architecture post describes the system behind both the app and its API. Every page cited by Perplexity AI passes through these six stages:

1. Crawl scheduling. A model decides when to index each URL, based on "the importance and likely update frequency" of that URL. The index tracks "over 200 billion unique URLs."
2. Storage priority. Documents kept in fast storage favor "authoritative domains" and "undercovered topics."
3. Parsing. Pages are split into "self-contained spans, each of which can be individually retrieved and ranked at query time." Sites heavy on lists and tables "may benefit from more formulaic parsing."
4. Retrieval. Keyword and semantic search run side by side and merge into one candidate set.
5. Filtering. Prefilters remove "clearly non-responsive or stale content."
6. Ranking. Fast scorers cut the set down, then cross-encoder rerankers make the final choice, at "both the document and sub-document levels." The rankers learn from the "millions of user requests" served each hour.

What the API docs add

Perplexity's developer docs show what a source looks like as data. Each search result carries an id, the canonical url, a snippet described as "excerpted text extracted from the page during search", a publish date and a last_updated date. The same page says a question can be split into "several reformulated queries," and offers filters for recency and for publish or update dates.

Two caveats. First, the Sonar chat API is no longer supported. Perplexity's docs say "Sonar Chat Completions support ended on September 27, 2026," that old requests are being reformulated as Agent API requests, and that new projects should use the Agent API. Second, Perplexity's API FAQ says the API uses "the same search system as the UI with differences in configuration," so API results won't match the app exactly.

Crawlers and blocked pages

Perplexity documents two bots on its crawler page. PerplexityBot surfaces and links sites in search results and "is not used to crawl content for AI foundation models." Perplexity-User visits a page when a user's question needs it, and "generally ignores robots.txt rules." Its help center adds that for a page blocked in robots.txt, Perplexity "may still index the domain, headline, and a brief factual summary." Our PerplexityBot user agent guide covers the strings, IP lists and rules.

Publishers and paid placement

Perplexity's Publishers' Program launched in July 2024 with revenue sharing tied to planned ads. Those ads didn't last. MacRumors reported in February 2026 that sponsored answers were phased out and executives had no plans to bring them back. The newer route is Comet Plus, which shares subscription revenue with partner publishers on "human visits, search citations, and agent actions." Digiday reported an 80% share and a $42.5 million initial pool. Perplexity doesn't say partner status changes which sources get cited, so don't plan on it.

Documented, observed or unpublished

Domain Authority vs Topical Relevance for Being Cited by Perplexity AI

Perplexity puts both words in one sentence. Its Instant Buy help article says: "Just as our answer engine ranks sources based on authority and relevance, we rank product listings using similar criteria." It doesn't publish how the two are weighed.

The architecture post shows where each one acts. Authority shows up early: a URL's "importance" helps set how often it's indexed, and authoritative domains get priority in fast storage. Relevance shows up late, when each passage is scored against the reformulated query. A trusted domain stays current in the index. A relevant passage wins the slot.

What the studies observed

Read together, the picture is consistent. A few giant sites take a big slice of Perplexity's attention across all topics. Inside any one niche, though, citations spread out more than on other engines, and pages that already rank in Google show up often. Authority helps you get into the pool. Relevance at the passage level decides which page gets cited by Perplexity AI.

One finding cuts against a common belief. Ahrefs found Perplexity's cited pages were younger than Google's results on average, yet still more than three years old. Fresh dates help when the topic changes. An old page that answers the question well can still be cited by Perplexity AI.

The Retrieval Ladder: Find Where Your Page Drops Out

We call this diagnostic the Retrieval Ladder: six rungs a page climbs on its way to being cited by Perplexity AI. Each rung maps to a stage Perplexity documents. A page has to clear every rung below to reach the next one, so you diagnose from the bottom up.

The API checks belong to a developer. Our guide to Perplexity SEO tools includes a short script that logs which results were retrieved and which were cited. Without a developer, run rungs 1, 5 and 6 by hand and infer the middle from what wins.

Read the ladder before you rewrite

It's tempting to start at rung 4 and rewrite the copy. Sometimes the page never reached rung 1. A rewrite can't fix a 403, and a perfect passage inside client-rendered HTML can't be cited by Perplexity AI if its crawler never sees the text. Check the bottom rungs first. They're quicker to fix and they block everything above.

If you want the one-page version of this work, our 30-day plan to get cited by Perplexity takes a single page through every rung, day by day.

A Reachability Check for Perplexity

Rung 1 fails silently. Nothing in Perplexity tells you a page was blocked, and a blocked page can't be cited by Perplexity AI however good it is. Run these five checks once, then after any CDN, firewall or robots.txt change.

1. robots.txt. Confirm no group blocks PerplexityBot on the paths you want cited. Perplexity says changes can take up to 24 hours to apply. Our robots.txt tester shows which line decides each URL.
2. Firewall and CDN. Perplexity's crawler docs advise a rule that matches the user agent and the published IP ranges. On Cloudflare, the 15 September 2026 changes moved AI bots into new controls, and Cloudflare Radar didn't list Perplexity as a verified bot on 1 October 2026. Our Cloudflare AI bot management guide shows which settings reach PerplexityBot and Perplexity-User.
3. Raw HTML. Fetch the page without running JavaScript and search for your key sentence. If it's missing, the crawler sees an empty page. Our guide to prerendering for AI crawlers covers the fix for each framework.
4. Logs. Look for PerplexityBot fetching the URL after you publish, and for Perplexity-User hits, which mean a live question pulled the page. Paste a day of logs into our AI crawler log analyzer, then verify heavy hitters against the IP list.
5. Status and speed. Perplexity-User fetches while a person waits. Timeouts, 429s and 5xx errors cost you that answer.

There's no shortcut around the crawler. Perplexity's help center and docs describe no URL submission form, and it isn't on IndexNow's list of participating engines as of October 2026. Discovery comes from its crawler and from links.

Worked Example: Tallyfold Climbs the Ladder

Tallyfold is a fictional invoicing and payments app for agencies. Its rivals, Brindlework and Kestrelyn, are fictional too. Every number below is illustrative.

Tallyfold picks eight buyer questions and runs each three times in Perplexity's incognito mode: 24 answers. Its own pages are cited by Perplexity AI in 7 of them (29%), and its name appears in 6 (25%). Then it finds the highest rung each question reached.

That's 2 + 1 + 2 + 1 + 2 = 8 questions. Three of them stall on the bottom two rungs, and their fixes are technical, not editorial. The writing fixes come after.

Six weeks later, Tallyfold reruns the same 24 answers. Its pages are cited in 13 (54%) and its name appears in 10 (42%). That looks like a 25-point gain in citations. With only 24 answers per round, though, the 95% margin of error on the difference is about ±27 points. The gain is promising, not proven. Tallyfold should run the panel twice more before it reports a win, as our guide to measuring GEO explains.

Where Rankbox Fits

Rankbox doesn't track citations or brand mentions in Perplexity, and it doesn't manage robots.txt or firewalls. Use the free tools above for reachability, and the AI Visibility Prompt Kit for a set of prompts to run by hand.

Rankbox helps with rungs 3 and 4: having a page worth being cited by Perplexity AI because it answers the sub-questions Perplexity searches for. Answer-Space Research maps the questions buyers ask Perplexity, ChatGPT and Google, with model-estimated volume, difficulty and intent. The Citation-Ready Writer researches the live web and writes 2,000 to 3,500-word articles with their sources cited. Articles reach your site through Rankbox's API. The Business plan is $49.50 a month with a 7-day trial: see pricing.