How to get cited by Perplexity

How Perplexity Actually Works

Perplexity answers a query through a multi-stage pipeline built around its own retrieval model, Sonar. Unlike a traditional search engine returning ten blue links, Perplexity's job is to synthesize an answer and attach citations to the specific claims it makes.

The pipeline runs, in order: query intent parsing, hybrid retrieval (combining classic keyword matching like BM25 with dense vector-embedding similarity), a multi-stage reranker that scores candidate pages on relevance, freshness, structure, and authority, then structured prompt assembly that feeds the surviving passages, pre-tagged with their source, into the model that writes the answer.

The consequence: a typical query retrieves on the order of ten candidate pages and only three to five earn an actual citation in the final answer. Every stage of the pipeline is a filter, ranking well on one signal (say, domain authority) doesn't help if the passage fails the freshness or structure filter earlier in the chain.

How PerplexityBot Crawls Your Site

Perplexity's declared crawler identifies itself asPerplexityBot/1.0 (full user-agent string:Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; PerplexityBot/1.0; +https://perplexity.ai/perplexitybot)). It respects robots.txt, and Perplexity publishes the crawler's IP ranges for verification. Critically, Perplexity states PerplexityBot is not used to train its underlying models, it exists purely to index content for citation in search answers. That means allowing PerplexityBot costs you nothing on the training-data front; it only makes you discoverable.

robots.txt directives

To allow full indexing, simply don't block the agent. To block Perplexity entirely:

User-agent: PerplexityBot
Disallow: /

Rule changes typically take up to 24 hours to take effect. You can scope a disallow to a single section (e.g. Disallow: /internal-docs/) rather than blocking the whole site.

The undeclared-crawler caveat

Worth knowing honestly: Cloudflare has documented Perplexity using undeclared, stealth crawlers that don't identify as PerplexityBot when a site blocks the declared agent, fetching pages on-demand to answer a live query even where robots.txt says no. If you actively want to keep content out of Perplexity, a robots.txt disallow is necessary but isn't airtight on its own; server-side blocking by verified IP range is the more reliable lever for genuinely sensitive content.

The Freshness Engine

Freshness is the signal that most separates Perplexity from ChatGPT and Claude. Perplexity doesn't just serve from a static index, every query triggers a live web search layered on top of a continuously-refreshed background index, so newly published or updated content can surface within hours, not weeks.

The recency window

Content published or meaningfully updated in the last 30 days gets a measurable citation boost. On fast-moving topics, new product launches, pricing changes, breaking developments in a category, that effective window compresses to 48-72 hours: whoever publishes the current number first tends to get cited until someone more current displaces them.

Crawl priority isn't equal

Perplexity's crawl scheduling is itself modeled: high-traffic, high-authority domains (major news outlets, government sites, established blogs) get recrawled within hours of publishing. Smaller or lower-visibility sites can wait longer before a fresh page enters the index at all. The practical implication for a smaller domain: a visibledateModified, a changelog, or an explicit "updated" line is a cheap, high-leverage signal that a page is current, use it honestly, and only when the page actually changed.

Because Perplexity does not publish a guaranteed indexing SLA, treat freshness as a probabilistic lever, not a promise: keep the page factually current and the timestamp honest, and the odds of citation shift in your favor over the following days.

What Actually Gets Cited

Independent studies of Perplexity's citation patterns, analyzing thousands of real answers, converge on a few consistent findings.

Reddit dominates, unevenly

Reddit shows up in a large share of Perplexity's citations, third-party analyses put the number anywhere from roughly a quarter to nearly half of answers depending on the study and query mix, well above any single publisher domain. Perplexity is the outlier among the major engines on this: ChatGPT and Google AI Overviews cite Reddit noticeably less often. If your category has active Reddit discussion, a thoughtful, non-promotional presence there is a real citation channel, not a distraction from it.

Citations are concentrated, but not in one place

Beyond Reddit, no single domain dominates, the top domain on any given topic rarely exceeds a few percent of total citations, and the long tail of thousands of smaller sites splits most of the remainder. That's the opening for a smaller, well-structured site: you don't need to out-authority an incumbent, you need to out-answer them on a specific query.

Structure and schema measurably move the needle

Schema markup (Article, FAQPage, HowTo, Person, Organization) gives Perplexity unambiguous entities to cite instead of forcing it to infer them from prose, and third-party schema studies show a meaningful lift in top-3 citation rates for schema-enabled pages over pages without it. Named authorship with real credentials compounds the effect, anauthor entity the pipeline can resolve to a real person reads as a stronger credibility signal than an anonymous byline or none at all.

The Perplexity Playbook

Crawlability (do this first)

• Confirm robots.txt doesn't disallow PerplexityBot, check it explicitly, don't assume a generic AI-bot block leaves it out.
• Verify crawler IPs against Perplexity's published range if you're auditing server logs for real visits vs. spoofed traffic.
• Keep the pages you want cited in your sitemap.xml and out of noindex.

Freshness (the highest-leverage lever)

• Update dateModified only when the page's substance actually changes, stale timestamps on unchanged content are easy for readers (and, over time, ranking systems) to catch.
• Refresh statistics, pricing, and examples on a real cadence, monthly for an evergreen guide, immediately for anything time-sensitive.
• Publish original, current data (a small survey, a benchmark, a dated pricing comparison), it's exactly the kind of content Perplexity's recency bias rewards over a rehash of last year's post.

Structure & credibility

• Lead every section with a direct, standalone answer before supporting detail, the reranker scores passages, not whole pages.
• Add Article schema (headline, datePublished, dateModified, author) and FAQPage schema where the content is genuinely Q&A.
• Use a real named author with a credentialed bio, not "Admin" or no byline.
• Where relevant, participate honestly in category discussion on Reddit, Perplexity's heaviest-cited source, by a wide margin.
• Track what's actually landing with Otterly AI (fastest feedback loop for Perplexity specifically) or Profound.

Implementation Checklist

Tick items off as you implement, this page remembers your progress on this device.

Frequently Asked Questions