Capture the answer before you do anything else
Run the query that produces the bad answer, copy the thread URL out of the address bar, and screenshot the whole answer. Perplexity threads are per session, so a re-run makes a new thread ID and can come back with different text. Close the tab without that URL and the report you file later has nothing to point at.
Then open the Sources list on that answer and copy every URL in it. That list is the actual work. Perplexity cites live web pages, so each entry is a page sitting on somebody's website, with a publisher who can change it.
Report the answer, then email support
Perplexity's help article How can I report incorrect or inaccurate answers? tells you to use the flag icon below the answer. There's no flag icon. We read the toolbar off a live answer on September 8, 2026: it's Copy, Share and Fork on one side, then Helpful, Not helpful and More actions, the three-dot button, on the other. Mark it Not helpful and open the three-dot menu.
Then email support@perplexity.ai, and treat the email as the real filing, because it's the one version of this that leaves you a record. The article names the only two things a report has to carry: URL to the query, and description of the error and expected result. Write what the answer says, write what it should say instead, and stop there.
Use their words for the problem. The issue types the article lists include Misinformation and Outdated information, and a false claim about a person is the first one. Put the request type in the subject line, because Perplexity publishes no legal, privacy or DMCA address of its own: answer errors, DMCA notices and privacy appeals all land in that same support inbox. No response time is published for this anywhere.
Fix the page Perplexity is citing
This is the step that moves an answer. Take the source list from step one and work it the way you'd work any other bad page: get it corrected at the publisher, delisted, or removed. When the cited URL changes, the citation dies at the next crawl. Everything else here is a request you file and wait on. This is the part with a real success rate.
If the page sits on a site you control, noindex it or block the crawler. Perplexity's crawler documentation names two user agents, PerplexityBot for search and linking and Perplexity-User for fetches a person triggers, and says it may take up to 24 hours for their systems to reflect a change.
A block isn't a removal, and Perplexity's own two documents disagree about how well it even holds. The help center says PerplexityBot won't index the text of a site that disallows it, then adds that if a page is blocked, we may still index the domain, headline, and a brief factual summary. The developer documentation says Perplexity-User generally ignores robots.txt rules because a person asked for the fetch. Both are live. Expect a blocked page to still get named.
Going the other way, a page you want cited has to stay readable. Perplexity recommends allowing PerplexityBot in robots.txt and permitting requests from its published IP ranges, listed at perplexity.com/perplexitybot.json and perplexity.com/perplexity-user.json. The firewall rule pairs the user agent with those ranges, and the documentation warns that firewall changes take their own time to propagate.
File the Data Privacy Request only if it fits
The form is at perplexity.typeform.com/datarequest, and it's the only form Perplexity's Privacy Notice names. Clear the cookie bar with Accept all or Reject all, press Start, then I understand, continue. on the screen headed IMPORTANT. It says four minutes.
Question one is required: your relationship to Perplexity. The options are Customer/User of Perplexity AI, Employee, Job applicant, On Behalf of a Client, Other (please specify), and two options for people who got marketing or recruiting mail. Filing for somebody else means picking On Behalf of a Client, and that's the answer that can put a proof-of-identity or power of attorney requirement in play.
Question two is required and it's the one to get right. Pick Restrict Processing of My Personal Data, which the form's own landing copy labels removal of content. Do not pick Erase my personal data. That copy says erasure deletes your account and all your data, and Perplexity's GDPR article confirms it can't reach specific threads, specific time periods, or anything short of deleting the whole account.
Don't file this form for anything you can do yourself. Perplexity's GDPR article lists deleting threads, spaces, files and your account as self serve, and says that if a request comes in for any of those, refusing it is within their rights. On timing, the form promises a response within one month of receipt. Under GDPR they can extend that by up to two further months, and they have to tell you inside the first thirty days if they do.
Set the date, then re-test the way a stranger would
Put the one-month date in your calendar the day you file. Perplexity publishes no case number or status tracker for a filed request, and the answer report has no timeline at all, so that calendar entry is your entire follow-up system.
Re-run the exact same prompt, logged out, across more than one session. Answers are generated fresh on every query, so one clean result proves nothing. If the old claim comes back on the second or third try, it isn't fixed, and a single good answer is not something to report to anybody as resolved.
If they deny the request, the appeal needs one exact word. The Privacy Notice says to contact them with the subject line Appeal. Capitalized, in the subject line, or it's just another email into the support inbox.
When none of that moves it
Go back to the sources and go harder. Perplexity offers no further internal escalation for content about a person. Correcting, delisting or removing the cited page is the lever that works, and the second one is publishing better pages about the same person so there's something accurate for the model to read instead.
In the EU or the UK there's an escalation the United States doesn't have. VeraSafe is Perplexity's appointed Article 27 representative and its data protection officer, reachable at experts@verasafe.com, and the Privacy Notice says that if you're not satisfied with the response, you may have a right to lodge a complaint with your local supervisory authority.
If you own the copyright in the material, the DMCA turns a request into an obligation. That policy lives in section 6.5 of Perplexity's terms of service, not at the DMCA link in their own footer, which returns Page Not Found. It requires all six listed elements, including a statement of accuracy under penalty of perjury and a signature, sent to support@perplexity.ai or to 2261 Market Street, Ste. 14441, San Francisco, CA 94114. It reaches copyright and nothing else.
What this does not do
None of it obliges Perplexity to remove a true statement about you. Their terms say the service may generate output containing incorrect, biased, or incomplete information, so inaccurate output is written into the contract as an expected characteristic rather than a defect. Removal here is discretionary, every time.
And getting something out of Perplexity changes nothing at the publisher. The page you never got corrected is still sitting on the web being read, which is why that source list from step one is the thing worth your afternoon.