Prepare PDF text before using ChatGPT.

A practical example

You want to rewrite a letter containing Camille Laurent’s name and email address. Import the PDF, check the extracted text and detected details, then manually add anything missing. The details you approve are replaced with placeholders.

1. Review all masked text, including indirectly identifying details.
2. Copy only this text into ChatGPT, Claude or your approved AI. Do not attach the original PDF if your intention is to keep its details private.
3. Paste the reply into the same PreMask AI session to restore recognized placeholders. Check names, facts and the meaning of the result.

Scanned PDF or selectable text

A scan needs optical character recognition. It runs locally but may misread a name, column or digit. Compare extracted text with the document before approving masking. A complex PDF may lose its reading order during extraction.

On Windows, a PDF is limited to 100 pages and a file to 30 MB; combined extracted text is limited to 120,000 UTF-16 units. Oversized documents are rejected. See the help page for each edition’s limits and formats.

When do you need a different tool?

If you need to deliver a PDF with information irreversibly removed, use a dedicated redaction tool and verify the final file. PreMask AI prepares text for AI and restores replies. It does not promise to remove hidden information from every layer of the original PDF.

If your organization prohibits sharing the document with an external AI, masking does not override that restriction.

Try the workflow

Interactive example without an account · Where does your data go? · Installation and formats