How do I get cited by Perplexity?

Publish dated, specific updates and make sure PerplexityBot can reach them. Perplexity weights recent pages and community discussion heavily, and research puts fresh pages at around 6 citations against 3.6 for stale ones. Allow PerplexityBot in robots.txt, show a real updated date that matches a real change, keep your sitemap current, and be present in the threads where your category is discussed.

What is the difference between PerplexityBot and Perplexity-User?

One builds the index and the other fetches for a person. PerplexityBot surfaces and links sites in Perplexity search, respects robots.txt and is not used to train foundation models, while Perplexity-User visits a page when somebody asks a question and generally ignores robots.txt because the fetch was user requested. Blocking PerplexityBot removes you from the index the live fetch chooses from.

Does updating the date on a page make Perplexity cite it more?

Not on its own. Freshness that helps is evidence of an actual change: a new figure, a new section or a dated correction, reflected in both the visible date and the structured data. A new date on unchanged content is a claim you cannot back, and it trains readers and engines alike to stop trusting your dates.

Does Perplexity send much traffic?

Much less than it reads. Across the sites I run, requests identifying as Perplexity-User came to 3,167 over two weeks while visits from Perplexity came to 53, about one visit for every sixty fetches. Those are declared user agents rather than verified addresses, but the gap is large enough that citations, not referral traffic, are the fairer way to judge Perplexity.

Perplexity Remembers This Week

¶ PLAYBOOKTHE SHORT MEMORY

01One engine rewards what you did on Tuesday

Almost everything yesterday's entry said about ChatGPT is wrong for Perplexity. Perplexity names brands in roughly 13% of its answers against ChatGPT's 0.6%, and the difference comes from two places: how heavily it weights recent pages, and how much it leans on community discussion, Reddit above all. In the research I rely on, fresh pages earn around 6 citations against 3.6 for stale ones. Perplexity asks what is true now. ChatGPT asks what has been true for a while.

Two crawlers, and only one of them asks permission

PerplexityBot builds the index Perplexity searches. Perplexity says it respects robots.txt and is not used to train foundation models. Perplexity-User is the fetch that happens when somebody asks a question, and because a person requested it, Perplexity says it generally ignores robots.txt. The order matters: blocking the first quietly removes you from the pool the second is choosing from, which is why naming the crawlers you welcome beats trusting a wildcard somebody wrote for scrapers. Block the index and the live fetch has nothing left to find.

The community half of that difference is the part most sites skip. A recent thread where somebody asked for alternatives in your category is exactly the kind of page a recency engine reaches for, because it is dated, opinionated and written by somebody who is not the vendor. You cannot publish that thread yourself. You can be the name that comes up in it, which is slower work than writing, and no update schedule replaces it.

¶ PLAYBOOKWHAT MY LOGS SAY

02What a recency engine looks like from the server

It went quiet when I did

On this blog, PerplexityBot made 32 requests across fifteen days of August, a stretch when a new entry went up every morning. In the first half of September I published one entry here, and it did not request a single page. Across the other sites on the same server, which kept shipping through those two weeks, requests identifying as PerplexityBot came to 6,251. That matches what I noticed a month in, that crawling here follows publishing days rather than the calendar, and Perplexity follows it more tightly than anything else in my logs.

Reads are not visits

The live fetcher is far busier than the referrals suggest. Over those same two weeks, requests identifying as Perplexity-User came to 3,167, while visits arriving from Perplexity itself came to 53. Those are declared names rather than verified addresses, so read them as a shape, not an audit, but the shape is hard to argue with. On my sites, Perplexity fetched a page about sixty times for every visitor it sent. The engine is consulting you constantly and forwarding almost nobody, which is worth knowing before you judge it by its traffic.

EXTRACTED — THE SENTENCE THIS ENTRY EXISTS FOR

Freshness is not a date on the page. It is evidence that something on the page changed.

¶ PLAYBOOKWHAT TO PUBLISH FOR IT

03Give it something new that is also true

Date the change, not the page

The cheap version of freshness is changing the date. It is also the version that teaches an engine to stop believing your dates. What counts is a visible, specific change: a new figure, a new section, a correction with the day it was made. Put the updated date where a reader can see it and in the structured data, and make sure the thing it describes actually happened. A new date on an old page is a claim you cannot back.

// THE SAME PAGE AT TWO AGES~~AVERAGE CITATIONS~~

Fresh

6.0

Stale

3.6

WHAT READS AS FRESH

A specific new number or a new sectionCounts

A correction, with the day it was madeCounts

An updated date on the page and in the schema, matching a real changeCounts

A new date, and nothing else changedA claim you cannot back

A research average, not a promise for your page. A fresh page earns about two thirds more.

Keep the instruction files as fresh as the articles

A recency engine can only reward what it can reach while it is still recent. The sitemap, the robots file and the other files crawlers obey should change the moment a page does, because a stale copy of an instruction file sits between you and every fetch. Then announce the change through a ping instead of waiting to be rediscovered.

The pages that suit a recency engine best are the ones built to change: a monthly figure, a changelog, a tracker, an explanation of how something works this quarter. A small information site I run publishes a price table every month with the edition date at the top, and within weeks of its first edition it was one of the five pages AI assistants fetched most on that site. Nothing about it is clever. Every month it genuinely says something that was not true the month before.

One honest limit before anybody rewrites their blog every Tuesday. Chasing recency rewards churn, and a site that republishes the same article weekly to look busy is serving one engine at the expense of the one with the long memory. The two entries are meant to be read together: durable definitions for ChatGPT on the pages that should never change, dated updates for Perplexity on the pages built to.

Tomorrow: the engine least likely to name anybody, and why that makes its recommendation worth the most.