Correcting Perplexity: Fix the Sources It Cites

Perplexity attaches numbered citations to its answers. Each number points to a whole source, not a highlighted sentence. That is the useful part: when Perplexity gets your pricing or founding year wrong, the pages it leaned on are usually listed beside the claim.

Two cautions before you act on that list.

• Cited is not the same as used. Across AI platforms, the set of pages an engine retrieves is often larger than the set it cites. A wrong fact can come from a page that was read but not credited.
• No citations can mean no retrieval. If an answer shows no sources, the model may be answering from what it learned in training. Fixing a web page does nothing for that answer until retrieval happens. As of October 2026, Perplexity's Agent API only searches the web when the developer turns search on, so tools built on it can return ungrounded answers.

If you have not yet confirmed which kind of error you have, start with what to check when Perplexity is wrong about your company, and use the difference between a hallucination and an outdated answer to name it. This page assumes you have a cited answer and a wrong fact, and you want to fix the sources.

Open every source behind the wrong claim. Put each one in a group, because each group has a different fix.

Also note what is missing. If Perplexity cites three directories and none of your pages, the engine either can't reach your site or found nothing quotable on it. That is a fix on your side too. The reasons engines favor thin or old pages are covered in why AI cites weak or outdated sources about your brand.

Your pages are the only sources you fully control. Work through them in this order.

1. State the fact once, plainly. Put the current price, the product name or the one-line definition in a sentence on a page Perplexity can fetch. Stating a specific fact gives an engine a sentence it can quote.
2. Remove contradictions. Check the pricing page, the about page, the footer, old blog posts and press releases. If two of your pages disagree, Perplexity can quote either one. Update or redirect the stale one.
3. Put the fact in the HTML. Vercel's crawler study found AI crawlers fetch JavaScript files but don't execute them, so content rendered only in the browser can be invisible to them. Independent re-testing in June 2026 found PerplexityBot does not render JavaScript. Load the critical facts server-side.
4. Check robots.txt for PerplexityBot. Perplexity's documentation says PerplexityBot is for search indexing and is not used to crawl content for AI foundation models. If you block it, you remove yourself from the index that answers come from.
5. Check the network layer. A firewall, CDN or bot-management rule can return a 403 to a crawler that robots.txt allows. Nothing in robots.txt reveals that. The AI Crawler Checker requests a page as each AI crawler and shows which ones your firewall turns away.

A minimal robots.txt entry that allows the search crawler looks like this:

User-agent: PerplexityBot
Allow: /

Don't confuse this with Perplexity-User. Perplexity documents that this user-triggered fetcher generally ignores robots.txt rules, so blocking it is not a reliable control in either direction.

Skip the tactics with no evidence behind them. Schema, llms.txt and rewriting pages in an "AI-friendly" style are not documented fixes for what Perplexity says.

When Perplexity cites other sites instead of yours, those sites are the route into the answer. The evidence shows a strong association there, not a guaranteed effect.

For each third-party page with a wrong fact:

• Find the owner. Directories and review sites usually have a business-owner claim or edit flow. Publications have an editor or corrections contact.
• Send one specific correction. Quote the wrong sentence, give the right fact, and link to the page on your site that states it.
• Prioritize by citation. Fix the pages Perplexity actually cites before chasing every mention on the web.
• Keep a log. Record the URL, the request date and the outcome so you can re-check later.

The outreach itself, with templates, is in asking third-party sites to fix wrong facts. Competitor comparison pages are a harder case, because the owner has no reason to help you. See what to do when a competitor's comparison page misrepresents you.

Some pages are written to steer AI answers: hidden instructions, text shown only to crawlers, or invented claims dressed as reviews. A 2026 study of LLM search agents (arXiv:2606.16821) found that once manipulated content sits in the results, it can shift what an agent endorses, and that vulnerability differed by model and by deployment setup. That study used a mock search API, so it shows what happens after manipulated content is retrieved, not how often it reaches live results.

What this means for you:

• Read the cited third-party pages for text that addresses an AI directly or makes claims about you with no source.
• If you find it, report it to the hosting site or platform, and keep copies with dates.
• Never answer it in kind. Hidden text, cloaking for AI crawlers and prompt injection are illegitimate, and they put your own pages at risk.

This is a defensive check. It protects your brand from being misrepresented, and it doesn't raise your visibility.

Be clear about which parts of this are documented.

Documented, as of October 2026:

• PerplexityBot indexes pages for search and is not used for foundation-model training.
• Perplexity-User generally ignores robots.txt.
• On the Agent API, web search is opt-in, and results carry fields such as URL, title, snippet, date and last_updated.
• Answers cite whole sources by number.

Not documented anywhere in our evidence:

• How Perplexity ranks or selects sources.
• Whether page dates or update frequency influence selection.
• A page-level snippet control for Perplexity. The data-nosnippet attribute is documented by Google and Bing, not Perplexity.
• A publisher channel for requesting answer corrections. For the general question, see whether you can ask an AI company to correct your information.

So the honest plan is: fix what's broken on your pages, remove access blocks, get third-party sources corrected, then re-check. Perplexity retrieves per query, so a fixed source helps only when it is retrieved again. For the same process across every engine, use the fix plan for AI getting your brand wrong.

How long does it take for Perplexity to show a corrected fact? Perplexity publishes no timeline. A fix on your page only helps once PerplexityBot has fetched the new version and the page is retrieved for the question. Third-party fixes depend on when the site owner updates. Re-ask the same question every week or two, note which sources appear, and track whether the corrected page replaces the wrong one. If nothing changes, re-check crawler access before rewriting content.

Does blocking PerplexityBot remove wrong answers about my brand? No. Blocking PerplexityBot removes your own pages from Perplexity's search index. It does nothing to third-party pages that carry the wrong fact. In practice it often makes errors worse, because Perplexity then relies only on other people's descriptions of you. Keep PerplexityBot allowed and fix the content instead.

Perplexity cites my page but still states the wrong fact. Why? A few possibilities. The cited page may still contain the old fact somewhere, such as a footer or an archived post. The correct fact may load only through JavaScript, which PerplexityBot does not execute. Or the wrong claim came from another page that was retrieved but not cited. Check the raw HTML of your page first, then review every other source in the answer.

Will adding schema or an llms.txt file correct Perplexity? There is no evidence that either changes what Perplexity says. No engine documents reading llms.txt, and Google says its Search ignores it. Ahrefs' matched test of 1,885 pages found adding schema produced no major uplift in AI citations. Existing markup does no harm, but the documented fixes are plain facts in the HTML, open crawler access and corrected third-party sources.

Does the same fix work for ChatGPT and Gemini? The source-level work carries over: your own pages, third-party pages and access checks matter everywhere. The crawlers and controls differ. ChatGPT search uses OAI-SearchBot, not PerplexityBot, and Google's AI features depend on Google's index and snippet eligibility. Each engine also answers some questions from training data with no live retrieval, which source fixes can't reach until a later model.