Why detector disputes escalate so fast
The client sees a number they believe is objective. You know you wrote it. Neither of you has anything to examine, so the disagreement becomes about trust, and trust disputes end contracts.
The way out is to introduce something checkable. A test that names its method, states its own limits and anchors itself to a specific file gives you both something to look at other than each other.
What to send a client
Your drafting history first: document version history, commits, or the dated outline you worked from. This is the strongest evidence you have and it costs nothing.
Then the evidence report, which states what was tested and, importantly, what the test cannot establish. Sending a document that admits its own limits reads as considerably more credible than one that claims to prove your innocence.
Keep the tone procedural. You are not disputing that they ran a tool; you are asking what the tool’s false positive rate is and what the contract says about acceptance criteria.
Getting ahead of it next time
Put an AI clause in the contract: what is permitted, what must be disclosed, and what evidence is acceptable if a dispute arises. Naming an agreed process before there is money at stake is far easier than negotiating one afterwards.
If you check drafts routinely, the Pro tier’s batch upload and API let you run a whole delivery rather than a file at a time.
Wherever this page describes a result: a detected mark is not proof of authorship, and an absent mark is not proof of human authorship. WatermarkRemoverPro's on-device rewrite can reduce detectable evidence but cannot guarantee defeating a vendor's undisclosed watermark, on any tier.