GUIDE
How to remove your information from AI answers
Nearly every removal attempt fails for the same reason: it aims at the assistant instead of the page the assistant read. Three layers hold your information, and they come off in order.
Layer one: the page itself
Whatever the assistant said started as a public URL. That page is the only layer where removal is permanent, and it is the only one where anyone is obliged to answer you.
If you control the page, delete the line or take the page down. If it is on a platform, use that platform's own removal path. If it is a data aggregator that republished your details, use their opt-out — they are the reason the same stale bio appears in five places at once.
Do not skip this to go argue with the assistant. As long as the page is up, every re-crawl puts the claim back.
Layer two: the search index
A removed page can keep answering for weeks because indexes cache it. Google's Results about you tool at myactivity.google.com/results-about-you handles personal contact details and can request removal of results that surface them.
The assistants have their own channels for personal-data requests — OpenAI at privacy.openai.com, Anthropic at privacy.anthropic.com. These are requests about data processing, not a delete button on an answer, and they are slower than fixing the source.
If you own the page and want it gone from indexes but kept online, noindex on the page and a Disallow in robots.txt do more, faster, than any request form.
Layer three: what the model already learned
Training data is the layer you cannot reach. A claim baked into the weights stays until a later model is trained without it, and no one can promise you a date.
This is the honest ceiling of every removal service, ours included. What you can do is make sure the live-search layer — the one that answers with citations — points at something accurate, so the correct version is what gets cited going forward.
What to expect, in time
Source page edited: minutes. Re-crawled: days to weeks, and only if the crawler is allowed in. Assistant answers change: after the re-crawl, unevenly, one assistant at a time.
So re-check on a schedule instead of refreshing the same chat. A removal is confirmed when the claim stops appearing with a citation — not when you send the email.
Can I make ChatGPT forget something about me?
Not in the training layer. You can get the underlying page removed and request that indexes drop it, which stops the claim from being retrieved and cited. The frozen training snapshot stays until a future model is trained.
Is there one form that removes me from all AI?
No. Each company handles personal-data requests separately, and none of them controls the pages that carry your information. The source page is the only removal that applies everywhere at once.
How long does removal take?
Days to weeks for the search layer once the source is gone, assuming crawlers can re-read the page. Anything promising same-day removal from an assistant is describing something that does not exist.
OpenProfiles prepares and tracks these requests. It never marks a source as corrected because a request was sent — only a later audit that no longer finds the claim counts as evidence.
What ChatGPT actually knows about you
Assistants answer from two different places: frozen training data and a live web search run at question time. Here is how to tell which one is talking, and how to test it without fooling yourself.
How to correct wrong information AI says about you
Most bad AI claims are not lies, they are stale facts. Correcting is a different job from removing, and it needs a better source rather than a deleted one.
A repeatable AI visibility check
A ten-minute audit you can run monthly on a person or a brand, and the four pieces of evidence worth writing down each time.
Run an email or handle through the sources and see what each one returns, with the URL it came from — audit screen.
Last reviewed 2026-09-01.