Hands off my PDF
Put an AI opt-out in your PDF
Add a machine-readable rights reservation to any PDF, free. AI firms that follow the W3C's TDM Reservation Protocol should keep your file out of training and data mining.
Stamp a PDF
This PDF is digitally signed. Stamping it here will invalidate the signature, so stamp the unsigned original and sign it again afterwards.
This is a PDF/A file. Once stamped, it may fail PDF/A validation, because the TDMRep fields aren't declared in its metadata schema.
A link to a TDMRep policy file (JSON) that explains how to license your content. Leave it blank and the reservation still applies.
Enter a full web address starting with https://
Stamped. Your download has started.
0 pages across 0 PDFs stamped so far
The small print
Your file never leaves your device
Everything happens in your browser. Your PDF is never uploaded, stored or logged, and there's no account to make. The only thing the page sends anywhere is a bare ping to the counter after each stamp, carrying the number of pages and nothing about you or your file. Turn off your Wi-Fi once this page has loaded and it still works. That's the proof.
What the stamp can and can't do
- It records, inside the file, that you reserve your rights to text and data mining. Under EU copyright law, AI firms relying on the mining exception have to respect that.
- It can't stop anyone reading the PDF. Paste it into a chatbot and the chatbot will read it.
- Support among AI firms is patchy. Use it alongside your copyright notice and terms, as a clear record of your wishes.
Under the bonnet
The tool adds two fields to the PDF's XMP metadata, under the namespace http://www.w3.org/ns/tdmrep/: tdm:reservation set to 1 (rights reserved), and tdm:policy with your policy URL, if you gave one. It also says the same thing in plain English where tools that have never heard of TDMRep will look: the standard XMP Rights fields (Marked, UsageTerms and, with a policy URL, WebStatement), Dublin Core rights if the file has none, and “no AI training” keywords in the document properties. Your existing metadata is kept, your own copyright line and terms are left alone, and any old TDMRep values are replaced. It runs on pdf-lib, an open-source PDF library served from this site.
Who built this?
Who do you work with?
I'll work with any startups, scale-ups, charities and established brands who are looking to improve how they tell their story and connect with their audience. If that's you, I can help.
