E11 Apps

Hide your personal details before sending a PDF to ChatGPT

By E11 Apps, the developer of PDF Burrow. Updated . 4 min read.

Good PDF privacy with AI comes down to one habit. Before you send a PDF to ChatGPT or any online AI chatbot, replace your personal details with neutral placeholders such as [Person_A] or [Account_1]. The chatbot can still explain the document, because it rarely needs your real name or account number to do so. When the answer comes back, you swap the real details back in on your own phone.

Why PDF privacy matters with AI chatbots

When you upload a document to an online AI service, its contents travel to that company's servers to be processed. What happens next depends on the service and your settings. Some services may keep conversations for a time, and some may use them to improve their models unless you opt out. Policies change, and they are not always easy to read.

For a public report, that hardly matters. For your payslip, bank statement, tenancy agreement or medical letter, it does. Those documents often contain:

  • your full name and address;
  • national ID or passport numbers;
  • phone numbers and email addresses;
  • bank account and card numbers;
  • salary, balances and other financial details;
  • names of family members or employers.

Good PDF privacy with AI means sharing what the chatbot needs to help, and nothing more.

What is pseudonymization?

Pseudonymization means replacing identifying details with consistent stand-ins, so the text still makes sense but no longer points to a real person. "Ali Khan paid 12,000 from account 0123456789" becomes "[Person_A] paid 12,000 from [Account_1]". The chatbot can still reason about who paid what, while the real name and number stay with you.

The key word is consistent. If "Ali Khan" is always [Person_A], the AI's answer will make sense, and you can map it back afterwards.

How to hide personal details before sending a PDF, step by step

  1. Decide what the chatbot actually needs. To explain a lease clause, it needs the clause, not your name.
  2. Make a copy of the PDF, or its text, so the original stays untouched.
  3. Replace each personal detail with a placeholder. Use the same placeholder every time the same detail appears.
  4. Look for details hiding in less obvious places: headers, footers, signatures, reference numbers and email addresses.
  5. Keep a private list of what each placeholder stands for.
  6. Send the cleaned version to the chatbot and ask your question.
  7. When the answer comes back, replace the placeholders with the real details on your own device.

What should you hide, and what can you leave?

Hide anything that identifies you or opens an account:

  • names, addresses, phone numbers and emails;
  • ID, passport, tax and national insurance style numbers;
  • bank, card and policy numbers;
  • employee or customer reference numbers.

You can often leave amounts, dates and general wording, since those are what the chatbot needs to help. If even the amounts are sensitive, round them or ask your question in general terms.

Why redaction alone is not quite the same

Redaction removes information permanently, usually as black boxes. It is the right choice for a document you are sending to another person. See how to redact a PDF properly.

For AI, placeholders work better. A black box tells the chatbot nothing, so its answer may be vague. "[Person_A] owes [Person_B]" keeps the meaning, and lets you restore the details in the answer.

Watch out for these leaks

  • Metadata. The author field may contain your name even if the pages do not. See how to change a PDF's title and author.
  • Scanned pages. If a page is an image, text-based tools cannot see the details on it. They need to be read with OCR first, or covered by hand.
  • Your question. It is easy to clean a document, then type "I'm Ali Khan and my account is…" in the chat. Keep the question anonymous too.
  • Screenshots. A screenshot of a page bypasses all of this. Clean the document, not just the file you upload.

Is it safer to use an on-device AI instead?

For many tasks, yes. An assistant that runs on your phone does not need to upload the document at all. It is usually smaller than the big online models, so it suits summaries and direct questions. When you want the bigger online model, placeholders are the safer route. Our guide on how to summarize a PDF and ask it questions compares the two approaches.

Hiding personal details with PDF Burrow

PDF Burrow has a Safe for AI tool that hides your name, CNIC, phone and account details before you send a PDF to ChatGPT, then puts them back in the answer. You paste the chatbot's reply into the app, and the real details are restored on your phone. The cleaning happens on the device.

Frequently asked questions

Does a chatbot need my real name to explain a document?

Almost never. It needs the wording and the numbers that matter to your question. Placeholders let it follow who is who without knowing your identity.

Can I just tell the chatbot not to save my data?

You can change the privacy settings the service offers, and that helps. It does not stop the document being sent to its servers in the first place, so removing details first is still sensible.

What if the document is a scan?

Text-based tools cannot see details in an image. Run OCR so the text can be read and replaced, then check the result by eye, as OCR can misread numbers.

Is the answer still accurate with placeholders?

Usually, because the meaning of the document is unchanged. Check that placeholders were restored correctly before you act on the answer.

More PDF Burrow guides