Why Is My Site Not Showing Up in Perplexity?

You already ran the ChatGPT version of this question. Now you are asking it about Perplexity, and you are assuming the answer is the same.

It usually is not. Perplexity fails you in different places, and the most common one is a door, not a document.

Perplexity is a different machine

ChatGPT with search on reads the live web through OpenAI's fetchers. Perplexity runs its own crawler and builds its own index, and it documents both jobs separately.

PerplexityBot is the indexer. Perplexity's crawler documentation describes it as "designed to surface and link websites in search results on Perplexity" and states plainly that "it is not used to crawl content for AI foundation models" (Perplexity crawlers, checked 2026-08-06).

Perplexity-User is the live fetch. It fires when someone asks a question that touches your page. The same page notes that "since a user requested the fetch, this fetcher generally ignores robots.txt rules."

Two agents, two purposes, two independent controls. Perplexity's own summary of its robots.txt tags says "each setting works independently, and it may take up to 24 hours for our systems to reflect changes."

So a site can be indexed and never live-fetched, or live-fetched and never indexed. Those look identical from your side and have opposite fixes.

The tell nobody reads

Here is the part of Perplexity's documentation that tells you where the real failures are.

A meaningful share of that page is not about crawling at all. It is step-by-step firewall configuration: how to build a Cloudflare custom rule combining user agent and IP source address with the action set to Allow, and how to create IP sets and string match conditions in AWS WAF with priority above your blocking rules.

Vendors do not write WAF tutorials for fun. They write them because the traffic keeps getting turned away by security infrastructure the site owner never configured on purpose.

That reframes the whole diagnosis. Before you rewrite a paragraph, find out whether the crawler is being let in.

The checklist, in the order that finds it fastest

1. PerplexityBot is disallowed in robots.txt

The cheapest thing to rule out. Open your robots.txt and look for PerplexityBot by name, and for a blanket rule that catches it.

This gets set by accident more than you would think, usually by a plugin's "block AI bots" toggle that treats the search indexer and the training crawler as one decision. They are not one decision, and Perplexity states its indexer is not used for foundation model training. A robots.txt written against GPTBot says nothing about PerplexityBot either; every vendor reads only its own tokens.

2. A firewall or CDN is returning 403

The one Perplexity documents around. Bot-fight settings, managed WAF rules, rate limits, and country blocks all refuse crawlers without ever touching robots.txt, and a crawler that honors robots.txt would have made the request anyway.

You cannot see this from your browser, because your browser is not sending the crawler's user agent. Perplexity publishes the full string, Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; PerplexityBot/1.0; +https://perplexity.ai/perplexitybot), precisely so you can test with it.

If you have server logs, this is where the answer lives. How to find AI crawlers in your server logs has the grep and the status-code tally.

3. You changed something yesterday

Perplexity states changes may take up to 24 hours to be reflected, and that is only the settings layer. Recrawl and reindex are further waits on top.

Nothing in that sequence is fast, and none of it is a reason to change three more things while you wait.

4. Nothing on the web points to the page

An index needs discovery. If a page has no internal links from anywhere prominent, is missing from your sitemap, and has no external link at all, a crawler has no route to it.

This is the quiet one for new sites. There is no permission problem and no content problem. There is no path.

5. The page has no sentence worth lifting

Perplexity writes an answer with numbered citations attached to specific claims. Its help center describes the loop as searching the internet in real time, gathering from sources, summarizing, then citing, with "each answer includes numbered citations linking to the original sources" (Perplexity help center, article last updated May 1, 2026, checked 2026-08-06).

A page that circles a topic gives that process nothing to attach a number to. A page that states the price, the definition, or the three options in plain text does. That is the whole craft behind writing an answer capsule an engine can extract.

6. The answer only exists after JavaScript runs

A crawler fetching raw HTML gets what the server sent. If your content is assembled client side, inside a tab, or behind an accordion, the fetch can come back with a shell.

Load the page with scripts disabled. If the answer disappears, so does your candidacy.

7. Something else already answers it better

Sometimes nothing on your side is broken. Perplexity retrieves many candidates and cites a handful, and a clearer, better structured page won the slot.

This is where the honest boundary sits. Items 1 through 6 are documented and checkable. Which page gets picked from the survivors is selection logic Perplexity has not published, and why some pages get mentioned in AI answers and others do not walks that line rather than pretending it is solved.

What does not transfer from your ChatGPT diagnosis

Three things, and each one costs people a week.

Your OpenAI robots.txt work does nothing here. Different vendor, different tokens.

Your ChatGPT result does not predict your Perplexity result, because ChatGPT with search reads the live web through OpenAI's fetchers while Perplexity ranks against an index it built earlier. Fresh pages and old pages behave differently across those two paths.

And a training-crawler block is not a search-crawler block. Perplexity says its indexer is not used for foundation models, so opting out of AI training and opting out of Perplexity's search results are separate choices that people routinely make as one.

Old way, new way

The old way was to treat "AI search" as a single destination and to fix it with a single change.

The new way is per engine, because the engines are per engine. Four different retrieval paths, four different sets of tokens, four different answers to the same question about the same URL. The list of AI answer engines runs longer than four, and each one repeats this story with its own tokens; Google's AI surfaces have their own failure modes too. Your position is not a rank. It is a set of yes-or-no reads, and how to check if your website shows up in AI search covers what each checking method can and cannot carry.

Verify before you report a number

Two verification steps that stop most wrong conclusions.

The user agent is self-reported, so anything can claim to be PerplexityBot. Perplexity publishes IP address files for both agents and tells you to treat them as the source of truth for WAF rules. Check the addresses before you write "Perplexity crawled us" in a report.

And check the status codes, not the hit count. A crawler that fetched your page and got a 403 looks, from inside an answer, exactly like a crawler that never came.

The damaging admission

Citedon's Perplexity check queries Perplexity's API using the sonar-pro model and reads the citations that come back. That is a real live-web retrieval by a real Perplexity model, and it tells you whether your page can be retrieved and named.

It is not a screenshot of the consumer perplexity.ai app for a logged-in user, and we do not claim it is. If a tool tells you it reproduces the exact consumer answer, ask which endpoint it queried.

We cannot make Perplexity cite you, and we will not sell that. Selection is theirs. We measure whether your page can be read, and we keep measuring as that changes.

The automated fix layer is WordPress only, applied through the connected plugin with a preview and a per-fix approval. On Shopify, Wix, Squarespace, Webflow, or a headless build, the scan still diagnoses the page and you apply the structure yourself.

And if items 5 and 6 are your problem, no markup rescues you. Structure makes a real answer legible. It cannot invent one.

Why this does not stay fixed

Perplexity's crawler list has grown and been renamed since it was first published. Your WAF vendor ships new managed rules without asking you. Your own site adds pages every month that never got the structure your fixed pages got.

Three moving parts, none of them frozen by a one-time audit. That is why readiness is a state you keep rather than a task you finish, and the mechanism behind it is described on the how it works page.

Where to start

Take the single page you would most want Perplexity to cite, and answer two questions about it in this order: can a crawler fetch it, and is there a quotable sentence in the first screen of text.

Run a free scan on that URL to see how ChatGPT, Perplexity, Gemini, and Claude read it today, and which structural pieces are missing. If the answer points at crawler access, allow or block AI crawlers in robots.txt is the next step, and the per-engine mechanics live on the Perplexity engine page.