How to remove personal information from a PDF before uploading it to ChatGPT
Before you upload a tax letter, bank statement, lease or medical bill to an AI, take out the parts that identify you: names, ID numbers, account numbers, addresses and dates of birth. Here are three ways, from free and manual to one click.
What to remove first
| Remove | Examples | Why |
|---|---|---|
| Government IDs | SIN, SSN, health card, driver's licence | Enough on its own for identity theft |
| Account numbers | Bank, card, policy, member numbers | Pairs with your name for fraud |
| Names | You, family, clients, employers | Ties the whole document to a person |
| Contact details | Address, phone, email, postal code | Locates you |
| Date of birth | Any format | A common identity check |
Usually keep amounts, dates of transactions and the document's wording. That's what the AI needs to help.
Method 1: copy the text and edit it (free, slow)
- Open the PDF, select all the text and copy it into a text editor.
- Replace each private detail with a tag that stays the same everywhere it appears, like [PERSON_1] or [ACCOUNT_1].
- Paste the edited text into the AI, not the original file.
- Swap the real values back into the answer yourself.
Scanned PDFs have no text to copy, so this method only works on digital ones.
Method 2: black-box redaction in a PDF editor
Paid PDF editors have a redact tool that permanently removes text under a black box. Drawing a black rectangle on top is not redaction: the text underneath can still be copied out. Use the real redact feature, save a new copy, and upload that.
The catch: the AI's answer will say "the account holder" and "the account ending in ███", and you lose track of who's who when there's more than one person.
Method 3: mask and restore automatically
A masking tool finds the private details for you, swaps them for consistent tags, and puts the real values back in the AI's answer. That keeps the answer specific without the AI ever seeing your identity. Stand-In does this on your own device, including scanned PDFs and photos of forms.
Do this in one step with Stand-In
Drop in a PDF, scan, Word or Excel file. Stand-In swaps every name, SIN/SSN, account number and address for a tag like [PERSON_1], then puts the real details back in the AI's answer. It runs on your device; nothing is uploaded.
Also: change one ChatGPT setting
On personal ChatGPT plans, go to Settings › Data controls and turn off "Improve the model for everyone" so new chats aren't used to train models. Temporary Chats aren't saved to your history, though OpenAI says they may still be kept for up to 30 days for safety review. Neither stops the text from being sent, which is why masking comes first.
Sources: OpenAI, Data controls FAQ