How to stop clients reusing your work with AI

1. Scope and delivery habits

Most reuse happens because the document that travels furthest has the most in it. Change what travels:

• Deliver results widely, method narrowly. The findings and the numbers are what the client paid for; the scoring model, the interview guide and the step-by-step approach are what they would reuse. Put the method in an appendix or a separate document and circulate it to fewer people — see protecting your methodology as a consultant.
• Send summaries for wide circulation. For a steering group, a board pack or “the wider team”, send a results summary; send the full document only to the people you are actually working with.
• Cloak drafts and methodology appendices by default. Drafts travel furthest and get the least care; the methodology appendix is the thing most worth reusing. Cloak both as a habit and decide about finals per engagement.
• Name copies per recipient. acme-review-v2-draft-jsmith.docx tells you later which copy went where. Keep a one-line log of who received what.

2. The contract

Habits raise the cost; the contract sets the rights. Two things to get right:

1. An AI-reuse clause. One or two sentences: the client will not use the deliverables to train, fine-tune or prompt an AI system to reproduce or reuse your methodology beyond the licensed scope, and will not remove protection from protected files. Sample language, and a narrower variant for nervous clients, is in An AI clause for freelance contracts and Telling your client.
2. Background IP and licence scope. If your terms assign everything to the client, an AI clause has little to attach to. Keep your frameworks and methods as background IP you own, license the deliverable for the agreed purpose, and say what “beyond scope” means (other regions, other business units, resale). That is the line a reuse conversation later turns on.

Say it out loud, too: one line in the delivery email (“drafts and methodology come lightly protected; plain copies on request”) turns the clause into a norm.

3. The file

Document cloaking adds hidden text to a .docx, .odt or .rtf that people never see but machine file-readers do: look-alike decoy values beside each real number, name, email and link; whole decoy paragraphs that some extractors prefer over the content Word shows; and hidden steering notes. Nothing you wrote is changed or removed. Dragged into an AI tool whose reader ingests hidden text, the file gives a noisy, unreliable read and the one-click “do the same for us” stops being free. Sunda Cloak does this on your device in one click.

It protects one path — the file itself, uploaded or attached to an AI tool. Copy-paste out of Word, screenshots, PDF export and extractors that strip hidden text all bypass it, and hidden is not secret: anyone who turns on “show hidden text” sees the decoys. Strength depends on which tool reads the file, and providers change their pipelines, so expect months, not years, from any technique. How the protection works lays this out in full; Protect a Word document from ChatGPT is the hands-on version.

4. Evidence

You cannot have a calm conversation about reuse without knowing which copy travelled. Two tools help, both with limits:

• A per-copy mark. Each cloaked copy carries an invisible mark of zero-width characters, designed to survive a copy-paste of the visible text, so a passage that turns up elsewhere can be traced to the copy it came from. It tells you which copy, not who. Occasionally it shifts a line or page break — proofread the cloaked copy like any deliverable.
• An optional tripwire. Turn it on and each cloaked copy carries one hidden reference link. If an AI tool (or a person's tool) actually fetches that link, the popup dashboard shows the time, the network address and the software that fetched it — not who, and not which AI product. Many tools never follow links inside uploaded files, so a hit is a clear signal and silence is no information at all.

Can you tell if someone used AI on your document? covers what detection can and cannot tell you.

5. What to do when you find out

Usually it arrives sideways: a slide deck from another region that reads like your appendix, or a tripwire hit. A sequence that keeps the relationship and the invoice:

1. Stay calm and gather evidence. Which copy (your naming and log), what was reused (side by side), and when. A tripwire hit shows a tool fetched the link from a particular copy; it does not show who or why.
2. Never accuse on an AI-detector result. Tools that claim to tell whether text was AI-written are unreliable, and reuse of your work is a different question from whether AI was involved. Evidence is the similarity and the copy trail, not a detector score.
3. Have the conversation. Assume good faith: “I noticed the framework from our engagement is being used for the other regions — great that it's useful. That's outside the scope we agreed, so let's talk about licensing it properly.”
4. Offer a licence or an extension of scope. Treat reuse as a sales opportunity first: an invoice for a broader licence or a roll-out engagement is usually a better opening move than a lawyer's letter.
5. Escalate only if you have to. If the client refuses and the value is real, take your contract and your evidence to a lawyer. This is where the clause and the licence scope earn their keep.

Frequently asked