Optimize for Perplexity: A Practical Checklist

Perplexity answers a question in stages, and only some of them involve your site.

1. Access. PerplexityBot fetches pages and stores them in Perplexity's search index. Perplexity's crawler documentation says PerplexityBot is for search indexing and is not used to crawl content for AI foundation models.
2. Retrieval. When a user asks something, Perplexity pulls candidate pages for that question. You don't see which queries it runs.
3. Generation and citation. The model writes the answer and cites some of the retrieved sources. Perplexity cites whole sources by a numbered index, rather than quoting a short span the way Claude's API does. Our page on whether AI engines cite whole pages or specific sentences covers the difference.

There is also a fourth stage that sits outside all of this: what the underlying model already knows from training, with no web access. Nothing you publish this quarter reaches that.

What this means: you can decide whether your page gets into stage 1. You can influence what's on the page if it reaches stage 3. You can't control stage 2 at all. The next move is to fix the stage you own.

Most Perplexity advice blurs these together. Here they are side by side, as of October 2026.

The left column is mechanical and checkable. The right column is where the outcome is decided.

The research backs this split. A critical survey of 45 GEO studies (arXiv:2607.14035, July 2026) concluded that already-retrieved content can change how it's cited, but no reviewed technique shows a stable, cross-platform causal effect on whether a page gets discovered. C-SEO Bench (NeurIPS 2025) found that most content-rewriting methods for AI citation were largely ineffective and often hurt the page's ranking in the model's context.

So work the left column first. It's the part where a fix actually changes something.

Robots.txt rules apply per bot. Allowing one AI crawler doesn't allow another, and blocking one doesn't block another.

The most common mistake is a blanket AI block. Someone adds a list of AI user agents to robots.txt to stop training use, and PerplexityBot goes in with the rest. That removes the site from Perplexity's index, even though PerplexityBot isn't a training crawler.

A minimal rule that allows it looks like this:

User-agent: PerplexityBot
Allow: /

Check three things:

• No Disallow for PerplexityBot on the paths that hold your product, pricing and docs pages.
• No wildcard group that catches it. If your file has User-agent: * with broad Disallow rules and no PerplexityBot group, those rules apply to it.
• Training and search bots kept separate. You can block training crawlers such as GPTBot or ClaudeBot and still allow PerplexityBot. The same logic applies to ChatGPT, covered in our ChatGPT search checklist.

One limit to know: Perplexity-User, the fetcher that runs when a user asks Perplexity to open a page, generally ignores robots.txt according to Perplexity's own documentation. Robots.txt controls the index. It doesn't reliably control answer-time fetches.

Robots.txt is a request, not a gate. The gate is your network layer: the CDN, the web application firewall, and any bot-management setting.

A firewall or CDN can block a crawler that robots.txt allows. When that happens, the crawler usually gets a 403 Forbidden or a JavaScript challenge page. Nothing in robots.txt or your analytics tells you. The site looks open, and the crawler sees a wall.

This got more likely recently. From 2026-09-15, Cloudflare set Training and Agent crawler categories to blocked by default on pages that display ads, for new domains, new sites on existing accounts, and existing free-tier customers. Search crawlers stay allowed by default under that change. If your site sits on Cloudflare, open its AI crawler settings and confirm how PerplexityBot is treated, rather than assuming.

How to test it:

1. Request a representative page, such as your pricing page, with the PerplexityBot user agent.
2. Record the status code. A 200 with real content is a pass. A 403, a 429, or a challenge page is a fail.
3. Confirm /robots.txt itself returns 200 to the crawler. A bot that can't read your robots.txt can't follow it.
4. If you have access, filter your CDN or server logs for verified PerplexityBot requests and look at the status codes it actually received.

The AI Crawler Checker requests a page as each AI crawler and shows which ones your firewall turns away. Our Perplexity visibility audit checklist lists the rest of the access checks in order.

This is the strongest evidence in the whole field. Vercel's crawler study found that ChatGPT and Claude crawlers fetch JavaScript files but don't execute them, so they can't read client-side rendered content. Independent testing in June 2026 found the same for PerplexityBot. Googlebot is the notable exception, which is why a page can look fine in Google and be empty to Perplexity.

What this means in practice: if your pricing table, plan names or product description load through JavaScript after the page opens, PerplexityBot may store a page with none of them.

How to check a page:

1. Fetch the raw HTML with JavaScript off (view source, or a plain HTTP request).
2. Load the same page in a browser.
3. Compare. Any price, plan, feature name or definition that appears only in the browser version is a gap.

The fix is server-side rendering or prerendering for the content that carries your facts. You don't need to re-architect the whole site. Start with pricing, product, comparison and about pages.

Also check content hidden behind tabs, accordions and "load more" buttons that fetch text on click. It has the same problem.

One honest limit: rendering decides whether your content exists in Perplexity's index. It doesn't make Perplexity more likely to choose you. It's the price of entry, not an advantage.

Once a page is reachable and readable, what's on it starts to matter. The evidence here is narrower, so keep expectations modest.

Put quotable facts in plain sentences

Stating a specific fact on a fetchable page gives the engine a sentence it can use. The 252,000-trial study "What Gets Cited" (arXiv:2605.25517) found that including explicit price information and a recent timestamp helped consistently, and that topical relevance and list position were the biggest drivers of being cited first. That study ran in a simulated setup with two candidate documents per trial, so treat it as a signal, not a guarantee.

In practice:

• Write the price as text: "Starter costs $49 a month", not a price inside an image.
• Give a one-line definition of what you sell near the top of the page.
• Use the same fact, worded the same way, everywhere it appears. If your pricing page and your about page disagree, an engine has two answers to pick from. The free check for contradicting facts on your own site finds those conflicts.

Say you run a hypothetical project-management tool. A page that says "Plans start at $12 per user per month, billed annually" gives Perplexity something specific. A page that says "affordable plans for every team" gives it nothing to quote.

Get named on the pages Perplexity already cites

For many buyer questions, Perplexity cites review sites, comparison articles and forums rather than vendor pages. Being named on those pages is associated with appearing in the answer, though no study measures a reliable effect. Brand prominence is strongly associated with citation too: a study of AI search citations (arXiv:2507.05301, data from March to May 2025) found citation concentration on Perplexity with a Gini coefficient of 0.77. That's a description of who wins, not a lever you can pull directly.

The practical move is to find out which third-party pages Perplexity cites for your category and whether you appear on them. Our playbook on finding the third-party pages AI engines cite walks through it.

These come up again and again in Perplexity advice, and the evidence doesn't support them.

• Blocking every AI bot to stop training. It catches PerplexityBot and removes your search eligibility. Block training crawlers by name instead.
• Trusting robots.txt alone. An Allow rule behind a firewall that returns 403 does nothing. Test the network layer.
• Adding schema or FAQ schema to get cited. Google says structured data isn't required for its generative AI features, and Ahrefs' matched test found AI Overviews citations fell 4.6% after pages added schema. Perplexity documents no use of it. Existing markup does no harm, but it isn't a lever for AI answers.
• Publishing an llms.txt. Google says its Search ignores the file, and no engine, Perplexity included, documents reading it.
• Rewriting pages in an "AI-friendly" style or chunking them. Google's AI optimization guide says you don't need to write in a specific way for generative AI search and that there is no ideal page length. C-SEO Bench found citation-oriented rewrites often hurt.
• Hiding text for AI crawlers. Hidden instructions, prompt injection and content shown only to bots are illegitimate. A June 2026 study (arXiv:2606.16821) showed manipulated content can sway AI search agents, and that is exactly why engines treat it as abuse. It puts your reputation at risk.
• Accidental snippet suppression. A nosnippet, max-snippet:0 or data-nosnippet inherited from a template can hide content from AI answers on Google and Bing. Check your templates before blaming Perplexity.

For content-writing questions beyond facts, see how to write a blog post for AI citations.

Perplexity doesn't offer a publisher dashboard that shows which of your pages it retrieved or cited. So any measurement is a sample.

What you can do:

• Re-run your access checks after every deploy, CDN change or robots.txt edit. Access breaks silently.
• Ask Perplexity the questions your buyers ask and note whether your brand appears, what it says, and which sources it cites. Results vary by plan and by the model a user's question routes to, so log out and repeat questions more than once.
• Check facts, not just presence. A mention with the wrong price is a problem to fix, not a win.

The free checker on this site runs up to three buyer questions and gives a snapshot of whether AI engines name you. It's a starting point, not continuous tracking. For the wider question of what's chance and what's control, read getting cited by Perplexity: control vs chance.

Does Perplexity-User follow robots.txt? Generally not. Perplexity's crawler documentation says Perplexity-User, the fetcher that runs when a user asks Perplexity to open a page, generally ignores robots.txt rules. PerplexityBot, which builds the search index, is the crawler that robots.txt controls. So robots.txt decides whether you're in Perplexity's index, but it isn't a reliable way to stop answer-time fetches. If you need to block those, the control sits at your firewall or CDN.

If I block GPTBot, does that affect Perplexity? No. Robots.txt rules apply to each user agent separately. GPTBot is OpenAI's training crawler and has nothing to do with Perplexity. Blocking it leaves PerplexityBot untouched, unless your robots.txt also blocks PerplexityBot or a wildcard group catches it. Read your file group by group and confirm PerplexityBot has the access you intend.

Should I add schema markup or an llms.txt file for Perplexity? Neither has evidence of getting you cited. Perplexity doesn't document reading llms.txt, and Google says its Search ignores the file. Ahrefs' matched test found AI Overviews citations fell 4.6% after pages added schema, and no study shows schema helps on Perplexity. If you already have markup, leaving it is harmless. Spend the effort on access and server-rendered facts instead.

How long does it take for Perplexity to pick up a fix? Nobody publishes a number. Perplexity doesn't document how often PerplexityBot recrawls a page, and no study measures the delay between fixing access and appearing in answers. What you can confirm right away is the fix itself: that PerplexityBot now gets a 200 response and that your key facts are in the raw HTML. After that, re-check answers periodically.

Does the Perplexity API show the same answers as the app? Not necessarily. As of October 2026, the free app uses an auto-router and paid plans let users pick from several models, so no single API model is the app. On Perplexity's Agent API, web search is opt-in, and code still on the retired Sonar endpoint can return answers with no web access at all. Treat API results as a proxy, not a copy of what users see.