How to get cited by Perplexity
Getting cited by Perplexity requires three things: allowing PerplexityBot in robots.txt and at the firewall, serving server-rendered HTML, and writing short, self-contained answers backed by named sources. Perplexity fetches live pages at query time and prints its sources as numbered links, so freshness and clean attribution decide inclusion.
What makes Perplexity different from other AI answer engines?
Perplexity is retrieval-first: every answer is assembled from live pages fetched at query time and listed as numbered, clickable citations beside the text.
Two consequences follow. Crawlability and freshness carry more weight than they do for an engine answering mostly from training data, and the citation is visible, so it can return a visit instead of dissolving anonymously into a summary. The second difference is less obvious and more useful: Perplexity picks pages, where the largest engines pick domains. In Previsible's July 2026 dataset, 13% of Perplexity's classified referral traffic landed on blog and long-form pages against 8.7% across the whole mix, while ChatGPT sent 28.8% of its traffic to sites' own internal search results - the signature of a model that trusts the domain but cannot name the page. The unit of optimization here is therefore the individual page, not the site.
Is Perplexity still worth optimizing for in 2026?
Yes, but as a quality channel rather than a volume one: Perplexity referral traffic has fallen 61% from the March 2025 peak.
Previsible's 2026 State of AI Discovery Report, built on 6.77 million LLM-referred sessions across 166 GA4 properties, records Perplexity peaking at 17,507 monthly referral sessions in March 2025 and dropping to 6,788 by May 2026. Over the same window Claude overtook Perplexity in March 2026, and ChatGPT took 92.4% of trackable standalone referrals. Three caveats belong with those numbers before anyone reallocates a budget:
- Referrals are not citations. The same report attributes part of the decline to Perplexity keeping users inside its own browser and agentic surfaces, where an answer can cite a page without ever sending a click. A falling referral count is compatible with flat or rising citation volume.
- One panel, one metric. 166 GA4 properties is a sample, not the market, and other 2026 studies place Perplexity's share of AI referrals higher. Read the trend as directional, not as a market total.
- The traffic that does arrive is well matched to long-form work. For a content-led brand, a Perplexity session is worth more than its share of the pie suggests.
The practical order stands: optimize for ChatGPT first, Perplexity next - and judge it by citations rather than sessions, because sessions now undercount it.
Which crawler does Perplexity use?
Perplexity runs two user agents: PerplexityBot builds and refreshes the search index, and Perplexity-User fetches individual pages during a live user request.
| User agent | What it does | Honors robots.txt? | Where you control it |
|---|---|---|---|
| PerplexityBot | Surfaces and links sites in Perplexity search results; documented as not used for foundation-model training | Yes - this is the setting that governs eligibility | robots.txt, plus a WAF or CDN allowlist |
| Perplexity-User | Visits a page to answer a live user question and links it in the response | Generally no - the fetch was user-initiated | WAF or CDN allowlist only |
Per Perplexity's crawler documentation, the two settings work independently and a robots.txt change can take up to 24 hours to register. Two practical notes follow. The company publishes current IP ranges for both agents as JSON endpoints, and documents Cloudflare and AWS rules that match on user agent and IP together. That is worth setting up: firewall and bot-management rules block far more sites than robots.txt ever does. Verification then belongs in your own logs, since Search Console reports Google's crawlers only. PerplexityBot activity shows up nowhere else.
How much does freshness matter for Perplexity citations?
More than for training-based models. Perplexity fetches sources at query time, so a stale page competes against whatever was published or updated most recently.
A page that was accurate in March and untouched since is not disqualified, but it loses to an equally good page revised last week on any question where recency is part of the intent. The working cadence is simple: revisit cited pages on a schedule, update the numbers and the dates together, and let the sitemap reflect real changes. Bumping a date without changing the content is worse than leaving it alone, because it teaches both search engines and answer engines to discount your dates.
What does it take to get cited by Perplexity?
A numbered citation takes four things: crawler access, server-rendered HTML, a short self-contained answer under each question heading, and a named source per claim.
- Allow PerplexityBot everywhere. robots.txt is the easy half; the other half is your CDN, WAF, or bot manager, where allowlisting the published IP ranges alongside the user agent prevents silent blocks. Confirm in server logs, not in theory.
- Server-render the page. AI retrieval crawlers read the HTML they are served and do not reliably execute JavaScript, so a client-rendered app leaves nothing to quote.
- Put the question in the heading. Phrase each H2 the way a user phrases the prompt, and answer it immediately underneath.
- Keep the answer capsule short and link-free. Roughly 20 to 25 words, no pronouns in the first sentence, elaboration and links in the paragraph below: the answer-first technique in full.
- Name a source for every number. The engine prints its sources, so a claim it can attribute cleanly is easier to lift than a confident assertion with nothing behind it. Authority is judged per claim here, not per domain: pair claims with sources.
- Give each question its own page. Selection happens at page level, not domain level, so one guide covering five topics competes badly against five focused pages.
- Stay indexable in Google and Bing. Retrieval draws on general web search as well as the engine's own index, so ordinary indexation still carries weight.
- Refresh on a cadence. Set a review interval per page and hold to it; on a live-retrieval engine, freshness is a ranking input rather than a hygiene chore.
Every step above has a technique guide behind it - browse the strategy cluster for the full set.
Most of this list transfers to Claude, which picks pages the same way and skews further toward long-form explanatory content. The access layer is where the two diverge: Anthropic splits its crawlers into three separately controllable agents rather than one, so an allowlist built for Perplexity does not cover it - see Claude vs Perplexity sourcing.
Automating Perplexity visibility
Crawler access, capsule-level rewriting, source hygiene, and a refresh cadence are recurring work rather than a one-time pass, and the engine mix keeps shifting underneath. Citematic can automate Perplexity visibility as a repeatable workflow and report Share of Model, so you can see which pages Perplexity actually cites and which have gone stale. For the engine with the largest measurable referral share, see how to get cited by ChatGPT.
Key takeaways
- PerplexityBot governs eligibility; Perplexity-User serves live user requests and generally ignores robots.txt, so the firewall matters as much as the robots file.
- Perplexity picks pages, not domains - one question per page beats one guide covering five.
- Perplexity referrals are down 61% from their 2025 peak, but referrals are not citations. To get cited by Perplexity and know that it worked, measure citations rather than sessions.
Frequently asked questions
How do you get cited by Perplexity?
Allow PerplexityBot in robots.txt and in your CDN or firewall, serve server-rendered HTML, and open every question-shaped heading with a short, link-free answer backed by a named source. Perplexity fetches pages at query time and prints them as numbered citations, so freshness and clean attribution decide whether a page gets used.
Which crawler does Perplexity use?
Perplexity runs two user agents. PerplexityBot indexes the web so pages can surface and be linked in search results, and the documentation states that it is not used to crawl content for training foundation models. Perplexity-User visits a page when a live user question calls for it. Allow both, and permit their published IP ranges at the firewall.
Does blocking PerplexityBot remove me from Perplexity answers?
Blocking PerplexityBot removes the page from Perplexity's index, so it can no longer be surfaced or linked as a numbered source. The documentation states that each robots.txt setting works independently and that a change can take up to 24 hours to register, so confirm the result in server logs rather than assuming it.
Does Perplexity-User respect robots.txt?
Generally no. Perplexity's crawler documentation states that because a user requested the fetch, Perplexity-User usually ignores robots.txt rules. Control over that agent therefore sits in the firewall or CDN, not in the robots file. PerplexityBot is the opposite case: it is the agent whose robots.txt setting governs citation eligibility.
Is Perplexity still worth optimizing for in 2026?
Yes, with realistic expectations. Perplexity's monthly referral sessions fell 61% between March 2025 and May 2026 in Previsible's 166-property panel, and Claude passed it in March 2026. But referrals are not citations, and the engine sends an above-average share of its traffic to blog and long-form pages, so it stays valuable for content-led brands.
Sources
- Perplexity, Perplexity Crawlers documentation. Checked July 26, 2026.
- Previsible, 2026 State of AI Discovery Report, published July 6, 2026: 6.77M LLM-referred sessions, 166 GA4 properties, November 2024 to May 2026. Checked July 26, 2026.
Back to AI citation strategies | Get cited by ChatGPT | Statistics with sources