Perplexity SEO: Check If It Cites You in 10 Minutes

Search for perplexity seo and the first page is mostly agencies selling you a Perplexity SEO service. Underneath them sit four guides, and all four give you roughly the same checklist. Write clear answers. Add schema. Get on Reddit. Build authority.

None of them start with the question that actually matters, which is whether Perplexity cites you today. Most people writing about this have never checked. It takes about ten minutes, and it changes what you do next, because the fix for “never crawled” and the fix for “crawled and ignored” have nothing in common.

So this post goes in that order. What Perplexity is doing under the hood, how to run the test on yourself, and only then what moves the needle. There is also a section on why nobody can honestly sell you Perplexity rank tracking, including us, which is the part I expect to annoy people.

What is Perplexity SEO?

Perplexity SEO is the work of getting your pages into the source list Perplexity shows under an answer, and getting your brand named in the answer itself.

Those two things are separate, and people constantly merge them. Being cited means Perplexity read your page to write the answer. Being named means the answer recommends you by name. You can have the first without the second, and when you do, you are supplying the research that sells somebody else.

There is no position one here. There is one answer, a handful of numbered citations, and a follow-up box. So the mental model from Google does not transfer cleanly, which is the honest half of the GEO versus SEO argument. What does transfer is almost everything about how pages get found in the first place.

How Perplexity actually finds sources

This is the part worth understanding properly, because it is where Perplexity differs from the other answer engines and it is why the generic AEO checklist under-serves it.

Perplexity started life on a third-party search feed. It does not run that way now. It crawls the web with its own bot and maintains its own index, and its own engineering write-up puts that index in the hundreds of billions of pages, with tens of thousands of index updates processed every second. That index is now sold as a product in its own right through the Perplexity Search API.

The sequence when someone asks a question looks like this. The query gets parsed into something searchable. The index returns a candidate pool. A reranking pass cuts that pool down on relevance, freshness and authority. Then a model writes an answer from the survivors and cites the pages it leaned on. Out of five to ten candidates, three or four typically make the citation list.

Two consequences fall out of that, and they are the practical ones.

Freshness carries real weight: Perplexity revisits different domains at different rates, with news-shaped sites crawled far more often than everything else. A page updated last month competes better than an identical page last touched in 2024. Dates on the page are not decoration here.

Being in the index is a prerequisite, not a given: Google will find you through links whether you like it or not. Perplexity’s coverage is its own, so a page nobody links to and no bot can fetch is simply absent from the candidate pool. It never gets to lose on relevance, because it never gets ranked at all.

Why crawler access is step one

Perplexity documents two agents, and the difference between them is not cosmetic.

So blocking PerplexityBot in robots.txt takes you out of the index that feeds most answers. That is the lever that matters, and it is one line in a file. The full user-agent strings, and the equivalents for every other assistant, are in our AI crawlers list.

Here is the bit people miss. Robots.txt is not the only place you can be blocked. Firewall rules, bot-fight settings and rate limiting all sit in front of it, and most of them were configured before any of these bots existed. Plenty of sites politely allow PerplexityBot in robots.txt while their CDN quietly returns a challenge page to it. Perplexity publishes IP ranges for both agents as JSON, so you can verify a fetch is genuine rather than guessing from a user-agent string that anyone can spoof.

There is also an uncomfortable footnote on this vein. In August 2025, Cloudflare published a report alleging that when its declared crawler was blocked, Perplexity fell back to an undeclared agent posing as Chrome on macOS, rotating IPs and networks, across tens of thousands of domains. Cloudflare removed Perplexity from its verified bot list and shipped rules to block the behavior. Perplexity disputed the characterization.

Take from it what you need for your own decision, which is this. If you want in, allow the bot properly and stop there. If you want out, robots.txt on its own has not historically been a guarantee, so back it with a network-level rule.

The ten-minute Perplexity test

Do this before you change a single page.

Write ten questions a buyer would actually ask. Not your keywords. Questions. “Best X for small teams”, “X vs Y”, “alternatives to Z”, “is X worth it”. Comparison and alternatives phrasings matter most, because those are the answers where a product gets named rather than explained.

Ask all ten and log two things per answer. Every domain in the citation list, and whether your brand appears in the written answer. Keep it in a spreadsheet. It is boring, and it is the whole point.

Sort what comes back into three buckets. Named in the answer, which is the win. Cited but not named, which means your page was trusted enough to read and your name still did not make it into the recommendation. And absent, which is either a crawl problem or a relevance problem, and the robots.txt check above tells you which.

The middle bucket is the useful one. We found the same pattern testing our own pages against another engine. One of our posts was read to build an answer that then recommended three competitors by name. The trust was already there. The mention was not. That is a content fix on a specific page, and it is a far better lead than “publish more”.

One rule while you do this. A single miss is not a verdict. These answers genuinely move. We have watched the same query on the same engine, two hours apart, cite nineteen sources and name nobody, then cite four and name us. So run the ten questions, then run the same ten questions two weeks later. The pattern is the finding. One run is an anecdote.

What moves citations, and what does not

Once you have the data, the fixes sort themselves into a short list. Most of it will look familiar, because most of it is SEO.

The first two rows carry most of the weight. Perplexity’s reranker is judging candidates on relevance and authority, which are the same signals classic search already measures. A page that ranks well for the underlying query is a page that keeps showing up in the candidate pool. That is why an agency selling you Perplexity SEO as a separate discipline is mostly selling you SEO with a coat of paint.

The llms.txt row will annoy someone. It is not that the file is harmful, it is that no major assistant has committed to reading it, and we wrote up the evidence on llms.txt separately rather than repeat it here. Ship it if you like. Do not put it above your robots.txt check.

The last row is the one people skip, because it is slow. Answer engines lean on corroboration, and a claim that appears on your site and nowhere else is a claim with one source. The mechanics of that selection are covered in how AI assistants choose citations.

Why nobody sells honest Perplexity tracking

Here is the take I would rather not have to write, given we sell an SEO product.

We do not measure Perplexity, and neither does anyone else, reliably. ContextBolt SEO can show you what ChatGPT and Gemini say about you, because that data is available at scale. Perplexity is not, and Claude is not. So we do not claim them.

Every Perplexity visibility score on the market is built the same way. The vendor picks a list of prompts, runs them on a schedule, counts how often your name appears, and turns that into a percentage. That is a survey, not a measurement. Two things break it. The prompt list is theirs and not your buyers’, so the sample is arbitrary. And the platform returns different answers to identical questions, which we have watched happen on our own domain inside two hours.

None of that makes the tools useless. A trend line built from a consistent sample still tells you whether you are going up or down, and the better AI visibility tools are honest about the method. It does make the number useless as a position. If a dashboard tells you that you are “34% visible in Perplexity”, ask which prompts, how many runs, and what the variance was. If the answer is vague, the number is decoration.

Your own ten questions, run by hand, twice, are worth more than a bought score. They are your buyers’ questions, and you saw the answers yourself.

How much should you actually care?

Less than the guides imply, and the number is worth having in front of you.

SE Ranking’s study of 101,574 websites, tracked over sixteen months to April 2026, puts Perplexity at 7.23% of AI referral traffic, behind ChatGPT at 74.78% and Gemini at 11.56%. Its US share of AI referrals actually fell across that window, from 11.42% to 6.85%, while total AI referral traffic grew 36%.

And AI referrals are still a small slice of all search referrals. So a Perplexity strategy as a standalone project is a bad use of a week for almost everyone reading this.

What is a good use of a week is the underlying work, because it pays out across all of them. Crawler access, pages that rank, direct answers near the top, real dates, mentions off your own domain. Do that once and you are optimizing for Perplexity, ChatGPT, Gemini and Google’s own AI features at the same time. That is the same conclusion the answer engine optimization guide reaches, and running the Perplexity test is what tells you whether it landed.

Where an agent fits in this

The two levers with the most weight here are ranking for the query behind the question, and knowing which questions are worth chasing at all. Both are keyword and SERP work, and both are miserable to do occasionally in a tool that costs $100+ a month.

That is what ContextBolt SEO is for. One MCP URL in Claude Code, Claude Desktop or Cursor, and you ask in plain language. What is the difficulty on this question phrasing, who ranks in the top ten for it right now, what does my Search Console already say about the queries I am close on. Ahrefs-grade data, $35 a month, and a 7-day free trial before you are charged anything.

On the AI side it covers what it can defend. It will tell you what ChatGPT and Gemini say about you, and which sources they read to say it. It will not tell you about Perplexity, because that data is not there to sell you honestly. For Perplexity, run your own ten questions. It is ten minutes, and it is more truthful than any score you can buy.