Perplexity SEO, without the guesswork

Perplexity SEO starts with two user agents, not with content

Perplexity uses two separate crawlers with two separate consequences: PerplexityBot builds the index it searches, and Perplexity-User fetches one page while somebody is mid-question — and blocking either one removes you from a different half of the product. That distinction is in Perplexity’s own crawler documentation, and it is the part of Perplexity SEO you can verify today with a text editor and a curl command. Everything after it — which sources get quoted, how often, in what order — is a measurement question, and we will say plainly where this page stops being measurement. We have not connected Perplexity. Zero calls, no data of our own. What follows is the vendor’s documentation, checked against their doc page, plus what we did measure on the two assistants we do run.

Both rows come from Perplexity’s crawler documentation, checked on 15 August 2026 and re-checked on 5 September 2026. The complete cross-vendor version is on the AI crawler page.

The robots.txt lines that decide it

If you want to be findable in Perplexity, these two rules are the floor:

User-agent: PerplexityBot
Allow: /

User-agent: Perplexity-User
Allow: /

Two things this does not do. It does not affect any other assistant — those are different tokens entirely. And it does not help if a firewall or bot-protection rule is returning 403 before robots.txt is ever consulted, which is the failure mode people miss because robots.txt looks correct while the crawler never gets a page. Send the request with the crawler’s own user agent and read the status code.

What we measured, and why it is not a Perplexity measurement

On 15 August 2026 we asked two assistants 12 questions each about one local market and kept every answer. ChatGPT attached 44 links from 25 domains; Claude attached 0 links while its transcripts show 277 pages read. The links ChatGPT showed went mostly to businesses’ own websites, with a short list of review platforms making up most of the rest.

None of that is a Perplexity number. It is reasonable to expect a citation-first product to behave more like the engine that cites than the one that does not, but expectation is not measurement, and a page that blurred the two would be doing exactly what this site was built to argue against. If you need a Perplexity figure today, the honest answer is that we cannot give you one; what we can give you is the ChatGPT run in full, with its method attached, so you can judge how far it travels.

What is worth doing anyway

1. Confirm both Perplexity crawlers get a 200. Not robots.txt in theory — an actual request with each user-agent string.
2. Put the facts in the server-rendered HTML. A crawler that does not execute JavaScript reads what the server sent: prices, services, hours, address, the answers to your common questions.
3. Answer the question in the first paragraph. Retrieval engines quote passages rather than pages, and a page that gets to the point in its opening lines is the one that can be lifted out of context intact.
4. Be present where the citations already go. In our own measured run that meant business sites and a handful of review platforms; work out that list for your own category rather than assuming ours.
5. Ask the questions yourself, and write down who gets named. Fifteen minutes in the product you are trying to appear in beats a tutorial. The question pack builder writes the set; Perplexity is one more place you can paste it.

What it would take for us to measure Perplexity

A direct API key. The model gateway we run through does not carry a Perplexity model at all, so the Perplexity leg of our tool has never made a real call — and until it does, we count our coverage as two assistants, not three. The front page keeps that list, and the audit includes a re-run of the same questions once the missing engines are wired up. Saying “we support five platforms” before that would be the kind of claim this whole site is a complaint about.

Perplexity questions

Where these numbers come from

• Perplexity — Perplexity Crawlers: the two crawler tokens, their user-agent strings and what each is for. Checked 15 August 2026, re-checked 5 September 2026.
• OpenAI — bots and user agents: the same search-versus-training split on another vendor, for comparison.
• Our own run of 15 August 2026: 12 ChatGPT answers and 12 Claude answers archived with full text, links, model version and timestamps. It contains no Perplexity data, which is why this page contains no Perplexity measurements.

Related

• Every AI crawler user agent, with vendor docs — the cross-vendor version of the table above
• What ChatGPT cites in its answers — the measured run this page refuses to extrapolate from
• The GEO audit rubric — crawler reachability, scored, with the rest of the sheet