How to Get Cited in Perplexity

Perplexity doesn't rank ten blue links and let you click through. It reads a handful of sources, writes an answer, and drops numbered citations next to the claims those sources support. That's a fundamentally different kind of visibility than a search engine listing, and it runs on different mechanics: a different retrieval loop, different crawlers, and a publisher program that genuinely does pay some sites for being cited.

Here's what's actually documented about how that works, and what to do with it. If you're new to this whole category of optimization, the GEO primer for beginners covers the broader principles this post applies specifically to Perplexity.

The retrieval loop behind every Perplexity answer

Perplexity runs on retrieval-augmented generation. For most queries, it performs a live web search, pulls back a set of current sources, and feeds those sources into a language model that writes the final answer with inline numbered citations pointing back to them. This is structurally close to what Claude and ChatGPT's search features do, but Perplexity was built around this loop from day one rather than adding it on top of an existing chat product, which shows up in how aggressively it searches even for queries that don't obviously need current information.

Perplexity's own developer documentation describes its Sonar API returning a search_results array alongside every answer, each entry carrying a title, URL, and date, plus numbered citation markers embedded directly in the response text that map back to specific entries in that array. If you want to see Perplexity's citation behavior in a form you can actually query and log in bulk rather than reading one answer at a time, that API is the way to do it. This mechanical overlap with how Claude's citation pipeline works, structured results plus inline citation markers, isn't a coincidence; both are retrieval-augmented systems solving the same underlying problem with different index infrastructure underneath.

Perplexity hasn't published its full ranking formula for which of the retrieved pages make it into the final cited set, and anyone telling you they've reverse-engineered the exact weighting is overselling it. What is documented, and what matters more for your actual to-do list, is the crawler infrastructure that determines whether you're even eligible to be retrieved in the first place.

It's also worth knowing that Perplexity runs more than one mode of search depending on how a user asks. A quick default query runs a lighter version of the retrieval loop. Perplexity's Pro Search and research-oriented modes run a deeper, more iterative version, issuing follow-up queries based on what the first round of results turned up before writing the final answer. That means the same page can be a candidate source at two different depths of the same product, and a page that's thorough enough to hold up under a follow-up query (because it actually addresses the edge cases, not just the headline question) has a better shot at surviving into the deeper research mode's citation list.

PerplexityBot and Perplexity-User are not the same bot

Perplexity runs two separate crawlers with two separate jobs, and conflating them is the single most common mistake in Perplexity-focused SEO advice.

PerplexityBot is the indexing crawler. It builds the index Perplexity draws on to surface and cite sites in search results, it respects robots.txt, and Perplexity's own documentation recommends explicitly allowing it if you want to show up in Perplexity's answers. Block it and you're opting out of citation eligibility entirely, the same way blocking Googlebot opts you out of Google's index.

Perplexity-User is a different animal. It fires in real time when an actual person asks Perplexity a question that requires fetching a specific page to answer accurately, and Perplexity's documentation is direct about this one: it "generally ignores robots.txt rules," because the request was triggered by a real user action rather than autonomous crawling. If you want to keep a specific page out of Perplexity entirely, blocking it in robots.txt won't reliably do that for Perplexity-User; you need firewall or access-control rules that match on user agent and IP range instead.

Both bots publish their IP ranges for verification, which matters if you're seeing traffic claiming to be one of these user agents and want to confirm it's actually Perplexity rather than a scraper spoofing the string. Perplexity's documentation also notes that robots.txt changes can take up to 24 hours to take effect on their side, so don't panic-check an hour after you edit the file.

A practical robots.txt entry for a site that wants both indexing and live citation eligibility looks like this:

User-agent: PerplexityBot
Allow: /

User-agent: Perplexity-User
Allow: /

Allowing both doesn't guarantee either one actually crawls or fetches your pages on any particular timeline, it just removes the one lever you fully control as a blocker. The rest comes down to content quality and whatever your CDN or WAF layer is doing above robots.txt, which is worth checking separately since a default "block AI bots" rule at the edge can silently override a permissive robots.txt file.

What actually gets you retrieved and cited

Once PerplexityBot can reach your content, the same core GEO principles that apply everywhere else apply here, with a few Perplexity-specific leans worth knowing.

Perplexity's interface renders a lot of its answers as structured output: numbered steps, short bulleted summaries, comparison-style breakdowns. Content that's already organized that way, with real numbered lists for processes and real tables for comparisons rather than walls of prose describing the same thing, maps more directly onto what the model is trying to produce. That doesn't mean turning every page into a bulleted list; it means that where your content is naturally a process or a comparison, structuring it as one on the page makes extraction cleaner.

Freshness carries real weight. Because every search result Perplexity retrieves includes a date, and because the product is built around answering with current information, a stale page on a topic that changes (pricing, availability, a regulatory threshold, a product's feature set) is working against itself in a way it might not on a search engine where an old page can still rank on accumulated authority. Update the page, change the visible date, and don't let "last updated 2023" sit at the top of a page about something that's moved on since.

Direct, self-contained answers near the heading still win. This isn't unique to Perplexity, but it's worth restating because Perplexity's citation style, numbered references tied to specific sentences, makes the cost of burying your answer more visible than it is elsewhere. If the sentence that actually answers the query is the fourth sentence in a paragraph, that's the one at risk of getting paraphrased into something less precise, or skipped for a competitor's cleaner sentence.

Named authorship and visible dates matter for the same trust reasons they matter to Claude and to human readers: an unsigned, undated claim competes poorly against the same claim with a byline and a recent timestamp attached, especially once a topic has three or four corroborating sources and the model is choosing between them.

Original data and direct reporting outperform another summary of what's already published. If your page is a repackaging of a stat that's already circulating on ten other sites, there's no structural reason for Perplexity to pick you over the other nine. If you ran your own survey, published your own numbers, or quoted someone directly, that's genuinely harder to find anywhere else, and it's the kind of content Perplexity has less choice about where to pull from.

Comparison and "versus" queries behave a little differently than single-fact questions, and it's worth planning for separately. Asking Perplexity to compare two products or services tends to pull in more sources than a single factual lookup, because the model needs enough material to speak to both sides fairly. If you only publish a page about your own product, you're one input into someone else's comparison and you don't control the framing. A page that does the comparison yourself, naming the competitor honestly and structuring the tradeoffs as a real table, gives Perplexity a single source that already does the synthesis work, which is exactly the kind of content a comparison query is built to retrieve.

What doesn't move the needle

A handful of tactics that show up in Perplexity-specific advice don't hold up against how the retrieval loop actually works.

Submitting your URL somewhere for "Perplexity indexing" isn't a real step. There's no submission portal; PerplexityBot crawls what robots.txt allows it to crawl, on its own schedule, the same as any other crawler. If you want faster inclusion, internal links from pages PerplexityBot already visits regularly, and a clean XML sitemap, do more than looking for a submission form that doesn't exist.

Stuffing a page with the literal phrase "according to Perplexity AI" or similar doesn't influence whether Perplexity cites it. The model isn't searching for pages that reference it by name; it's searching for pages that answer the user's actual question.

Treating Comet Plus enrollment as a prerequisite for general citation is a common and costly misreading. The two are separate systems: citation eligibility runs through PerplexityBot and content quality, open to any site; Comet Plus is a revenue-sharing layer for a specific category of news publisher. A product or business blog that isn't eligible for Comet Plus can still be cited constantly in regular answers.

Relying on social proof or backlink volume the way you might for classic SEO authority doesn't transfer directly. Perplexity's retrieval is query-time and content-driven; a page with ten thousand backlinks but a buried, vague answer loses to a page with no notable backlink profile and a clean, direct one, for that specific query.

Comet Plus: the part of this that actually pays

Perplexity runs a publisher program that's worth understanding separately from general citation eligibility, because it's a different thing with a different bar for entry.

Comet Plus launched in 2025 as a $5 subscription tier bundled into Perplexity Pro and Max, giving subscribers access to participating publishers' content through the product. Perplexity's own announcement and Search Engine Journal's coverage of the launch both describe an 80/20 revenue split, with 80% of that subscription revenue going to participating publishers and Perplexity retaining 20% to cover compute and platform costs. By early 2026, reporting put the program at a $42.5 million publisher payout pool with over 2,400 publisher partners enrolled, including major outlets like Condé Nast's titles, CNN, The Washington Post, and Fortune.

Publishers reportedly earn across three kinds of usage: human visits to their content through the platform, Perplexity citing them in an AI-generated answer, and Perplexity's agent features reading their content to complete a task for a user. What Perplexity hasn't disclosed is the exact weighting between those three, or what an individual payout actually looks like for a mid-sized publisher versus a major outlet, so treat the headline revenue-share number as a program-level figure rather than a per-article estimate you can budget against.

This program is aimed at news and media publishers with licensable editorial content, not every blog or product site, and the stated path to joining is contacting Perplexity directly about partnership rather than a self-serve signup. If you're not a news publisher, general citation eligibility (the PerplexityBot and content-structure work above) is the lever that applies to you; Comet Plus is a separate, narrower program on top of that for a specific category of publisher.

It's also a program that's still young enough to keep moving. The launch partner list and payout pool figure are both things Perplexity has updated publicly since the initial announcement, so a mid-sized publisher evaluating whether to pursue it is better off emailing Perplexity directly for current terms than working from a number quoted in an article from a few months earlier, this one included.

To track whether these changes actually move your Perplexity citations, see our comparison of Perplexity rank trackers, including which tools log the cited sources and not just mentions.

Tracking whether Perplexity is actually citing you

Ask Perplexity directly, in a fresh session or logged out, using the actual questions your customers type rather than your brand name in isolation. Brand-name queries are more likely to get answered from general knowledge without triggering a fresh search; specific, current questions are more likely to force the retrieval loop where citations appear.

If you have developer access, query the Sonar API with your target prompts and log the search_results array programmatically. This is the repeatable version of the manual check above, and it's what's actually powering the third-party AI visibility trackers that do this at scale.

Watch your server logs for PerplexityBot and Perplexity-User hits, cross-referenced against the published IP ranges so you're not crediting a spoofed user agent. A PerplexityBot crawl tells you you're indexed. A Perplexity-User hit on a specific URL tells you a real person's question just triggered a live fetch of that exact page, which is as close to real-time evidence of relevance as you'll get without a direct citation confirmed in front of you.

Third-party tools built specifically for this, Otterly, Profound, and Scrunch AI among them, track citation presence across Perplexity alongside ChatGPT, Gemini, and Claude on a recurring schedule, which is the practical option if you need this monitored continuously rather than spot-checked.

Does PerplexityBot respect robots.txt?

Yes. Perplexity's documentation recommends explicitly allowing it if you want to be eligible for citation in Perplexity's answers.

Does Perplexity-User respect robots.txt?

No, generally not. It's triggered by a real user's question in real time, and Perplexity's own documentation states it generally ignores robots.txt rules. Use firewall-level access controls, matched against Perplexity's published IP ranges, if you need to block it specifically.

Is Comet Plus open to any website?

No. It's aimed at news and media publishers with licensable editorial content, and the path to participation is direct contact with Perplexity rather than self-serve enrollment. Most product and business sites should focus on general citation eligibility instead.

Can I see exactly which of my pages Perplexity has cited?

Not through a dashboard Perplexity provides directly. The practical options are logging Sonar API citation data for your own target queries, checking server logs for crawler and user-fetch activity, or using a third-party AI visibility tracker that automates the same checks across a recurring prompt set.

How is this different from ranking in Google?

Google still returns a ranked list of links for the user to choose from. Perplexity synthesizes one answer and cites a small number of sources inline, so the competition isn't for position ten through one on a results page, it's for one of a handful of citation slots in a single generated paragraph. The content fundamentals overlap with SEO; the competitive structure doesn't.

Does Perplexity use the same approach for every query?

No. A quick factual question runs a lighter retrieval pass. Pro Search and research-oriented modes run a more iterative loop, following up on initial results before writing a final answer, which means thorough pages that hold up to a follow-up question have an advantage in those deeper modes specifically.

Start with the crawler check, not the content audit. Confirm PerplexityBot and Perplexity-User are actually allowed to reach your site, in robots.txt and in whatever WAF or CDN rules sit in front of it, before you spend an afternoon rewriting headings for a bot that's currently getting a 403. Once that's confirmed, the single edit that moves the needle most on an existing page is almost always the same one: find the sentence that actually answers the question your page is supposed to answer, and move it to the first two sentences under the heading that matches it.

The crawler and content principles here are specific to Perplexity, but they're one piece of a bigger picture. The AI search category covers the same ground for Claude, ChatGPT, and Google AI Overviews, and it's worth checking each engine's crawler policy separately rather than assuming one robots.txt decision covers all of them. If you'd rather have this kind of breakdown land in your inbox as each engine updates its approach, the newsletter is the easiest way to keep up without re-checking every vendor's docs yourself.