How to get cited by Perplexity: a playbook for brands

How Perplexity differs from ChatGPT

Perplexity is an answer engine built around search. When you ask it something, it searches the web, reads a small set of results, writes an answer from them and attaches a numbered citation to each claim. The sources appear above or below the answer, and you can click each one. That happens for nearly every question, including simple ones that ChatGPT would answer from memory.

ChatGPT is a model first and a search tool second. It decides per question whether to search, and for many questions it answers from training memory without citing anything. We cover that in how ChatGPT decides which pages to cite. For a brand, the difference looks like this:

The practical consequence: with Perplexity you are competing for a place in a short list of search results, question by question.

What is known about how Perplexity finds pages

Perplexity's own documentation describes two kinds of visits to a website, and the distinction matters for how you treat them in robots.txt.

• PerplexityBot is the crawler that builds Perplexity's search index. It visits pages on its own schedule so they can be found when a question comes in. If you block it, your pages can drop out of that index.
• Perplexity-User is the fetcher that reads a page live when a user's question needs it, for example when someone pastes a link or when the answer engine wants the current version of a page. Perplexity states that this fetcher generally does not follow robots.txt rules because it acts on a user's request, in the same way a browser does.

At the time of writing, Perplexity also uses results from other search providers alongside its own index, and the exact mix is not public. The safe assumption is that a page which ranks in ordinary web search for a question has a good chance of being read by Perplexity for that question, and a page that ranks nowhere is unlikely to be read at all. Our guide to AI crawlers and robots.txt explains the crawler names across every engine and how to see which ones visit your site.

Two more things follow from a search-first design:

1. Freshness counts. Perplexity tends to prefer recently updated pages for questions where recency matters (pricing, "best tools in 2026", anything with a year in it).
2. The answer is built from a few pages. Being the eleventh best page for a question is roughly the same as not existing.

The playbook: pages Perplexity likes to cite

Answer one question, and put the answer first

Perplexity's model reads your page looking for the sentence that answers the user's question. If that sentence is in the first paragraph, under a heading that matches the question, it is easy to quote. If the answer is spread across three sections, the model picks another page.

A worked example. Fernwick, a made-up scheduling tool, wants to be cited for "how do I let customers book a meeting without back and forth". A product page titled "Scheduling, reinvented" gives Perplexity nothing to quote. A page titled "How to let customers book a meeting without the back and forth" that opens with "Share a booking link that shows only the times you are free; the customer picks one and the meeting lands in both calendars" is a quotable answer in one sentence. The second page can be cited. The first cannot.

Rule: every page you want cited should pass the "first paragraph test". Read only the title and the first paragraph. Does it answer a question a customer would type?

Show a date, an author and a company

Perplexity shows a date and a source name next to each citation, and the model appears to weigh them. Undated, anonymous pages are at a disadvantage for anything time-sensitive.

• Put a "Last updated" date near the top of the page, and keep it honest: update the date only when the content changes.
• Name the author with a short line on who they are. For a company page, name the company plainly in the first screen.
• Add Article or Organization structured data with the same date and author, so the facts on the page and the facts in the markup agree. Structured data helps here only when it matches the visible page.

Rank in web search first

Because Perplexity reads from search results, your SEO work is your Perplexity work. The pages most likely to be cited are the ones already on page one for the question, or close to it. If a page you care about ranks nowhere, fix that before you polish the copy: internal links to it, a title that matches the query, a few external links, and a sitemap entry that is actually crawled. Track rankings and Perplexity citations for the same question and you will usually see them move together.

Be present on the third-party pages it prefers

For recommendation questions ("best scheduling tool for agencies", "Fernwick vs Allsite"), Perplexity rarely cites only the brand's own site. It cites review sites, comparison articles, forum threads and documentation, because those pages compare several options and carry other people's experience. Your own site can win the factual questions about you; the recommendation questions are decided on other people's pages. What to do about it, without spamming:

• Make sure your listing on the review sites that matter in your category is complete, current and accurate.
• Publish the comparison pages people actually search for, in your own voice, with real differences. Comparison pages are cited for comparison questions.
• Answer real threads as yourself, disclosed, where you have something useful to add. Why AI answers lean on Reddit, reviews and forums goes deeper.
• Keep your documentation public and crawlable. Documentation pages are quoted for "how do I" questions far more often than marketing pages are.

Be specific and quotable

A model citing a source needs a sentence it can lift. Compare "Fernwick offers flexible pricing for teams of all sizes" with "Fernwick costs 8 euros per user per month on the Team plan, billed yearly, with a free plan for one calendar". The second is a fact, attributable to a page with a date. That is what gets cited. Specific also means concrete nouns: the feature's name, the integration's name, the limit, the region. Each one is a potential match for a question someone types.

Keep the page readable by a machine

Perplexity reads the HTML it fetches. Content that only appears after JavaScript runs, content behind a login or a cookie wall, or a page that returns a bot challenge to crawlers, is invisible. Check the page with JavaScript switched off and see whether the answer is still in the text.

How to check whether you are cited

You need a list of questions and twenty minutes a month.

1. Write down 15 to 30 questions your customers ask, in their words, across the stages of buying: "what is", "how to", "best tools for", "X vs Y", "is X worth it". How to choose the prompts to track shows how to build that list.
2. Ask each question in Perplexity, logged out or in a fresh thread so previous context does not shape the answer.
3. Note the numbered sources. For each answer, record which sources are yours, which belong to competitors, and which are third-party pages (reviews, forums, press). Note the position of your citation too, since the first source is read more than the sixth.
4. Record whether your brand is named in the answer text, separately from whether your page is cited. The two differ: you can be named from a competitor's comparison page, or cited for a fact without being recommended.
5. Repeat monthly with the same questions, and compare. Answers vary between runs, so look at the trend over three or four months rather than a single week.

A spreadsheet with one row per question and one column per month is enough to start. In Echo this lives on the Prompts page, where the same questions run on a schedule across every engine, and the Citations page groups the cited pages by site.

Common reasons a brand never gets cited

• The pages don't rank. The most common reason by far. If your page is not in the search results Perplexity reads, nothing on the page matters.
• The crawler is blocked. A robots.txt line that blocks PerplexityBot, or a bot-protection rule that serves a challenge page to crawlers, keeps you out of the index. Check both.
• The answer is not on the page. Product pages written as slogans give the model nothing to quote.
• No dates, no names. Anonymous, undated pages lose to pages with a byline and a date for anything time-sensitive.
• The only presence is your own site. For recommendation questions, a brand that appears on no review site, no comparison article and no forum thread is not in the pool.
• The content is rendered by JavaScript. The crawler sees an empty shell.
• The facts are wrong elsewhere. If review sites or old press describe you incorrectly, Perplexity may cite them and repeat the mistake. Trace the wrong fact to its page and ask for the correction.