What to keep if a US or UK publisher asks about AI use

Last checked 12 November 2026. General information, not legal advice. TimeBinder does not detect AI, and a dated record is not proof of authorship, ownership or originality.

Why publishers ask

Standard publishing contracts include a warranty that the work is original to the author. The Authors Guild’s guidance says AI-generated text is not original human authorship for that purpose, so including it may breach the warranty and, in some cases, let a publisher end the agreement. It tells authors to ensure their AI use does not breach the warranties in their publishing contract, to tell their publisher about AI-generated uses so it can prepare an accurate copyright application, and says authors remain responsible for the manuscript they deliver. The UK Society of Authors advises members to seek individual advice and to have their contracts checked.

These are advocacy bodies’ positions, not law, and contracts differ. The Authors Guild’s model clause, for example, would have the author disclose any AI-generated text in the submitted manuscript and limit it to a small amount, and would stop a publisher requiring the author to use generative AI. Your contract may say something different, or nothing.

Step 1: read what your contract and policy actually say

• The warranty of originality and any indemnity.
• Any AI clause: disclosure duty, a cap on AI-generated text, a ban on AI use, or permission with notice.
• What your publisher’s submission guidelines or an agent’s email say about AI.
• If you self-publish: the platform’s policy. Amazon KDP, for example, asks you to tell it about AI-generated text, images and translations when you publish, but not AI-assisted content. It defines AI-generated as content an AI tool created, “even if you applied substantial edits afterwards”, and AI-assisted as content you created and improved with AI, or ideas you brainstormed with AI before writing it yourself. KDP also says you remain responsible for checking everything against its guidelines and intellectual property rights.

If anything is unclear, ask your agent or editor in writing, and keep the reply. Any rule you rely on should be in the contract or an amendment, not an informal assurance. A literary agent or a lawyer who knows publishing contracts is the right person for the legal reading.

Step 2: keep a short AI-use log

Whatever the rule, you will find it easier to answer if you wrote it down at the time. A one-page log is enough:

• The tool and version, and the dates you used it.
• What you used it for: spell-check, grammar, brainstorming names, outlining, a translation, research summaries, or generating text.
• Whether any of its output, in whole or in part, went into the manuscript, and where.
• Which services you gave your text to, and whether their settings allow training on it. The Authors Guild warns that material entered into public chatbots may be used for training by default, so check before uploading drafts.

Be honest in the log. A tidy log that is later contradicted does more harm than a messy one that is accurate.

Step 3: keep dated drafts, not just the final file

Publishers who ask about AI usually care about how the text came to exist. Dated drafts show the development of the work over time: outline, early chapters, revisions, edit letters, tracked changes from your editor. Save each as its own file, not an overwrite, and keep your notes and research alongside.

Group the drafts at milestones (first complete draft, post-edit draft, submitted manuscript) and create a TimeBind for each group: the fingerprints of up to 10 files are anchored on the Bitcoin blockchain, and the files remain on your machine. See draft history and manuscript timestamping. For a full writing routine see a writer’s manuscript workflow.

Step 4: what to tell them

• Say what you did in plain words, matching your log. Do not overstate or understate.
• Offer your dated drafts if they want to see development, and keep the originals unedited.
• Ask what they need in writing, and by when.
• Do not send private files or credentials; send what is asked for, or ask your agent.

If the question comes as an accusation rather than a routine request, slow down. See how to reply to an AI accusation and author accused of using AI.

What a dated record shows, and what it does not

A timestamp shows that an exact file existed no later than a recorded time and is unchanged. Taken with a draft history and a log, it can support your description of how you worked. It does not show that the text is human-written, that you wrote every word, or that the work is original, and it does not tell a publisher whether your use of AI is permitted. TimeBinder does not detect AI use; those questions are for your contract and your publisher. Detector scores are not a substitute either; see why working drafts and timestamps matter more than AI detectors.

US copyright has its own AI disclosure expectations when you register a work. See do you have to disclose AI on a US copyright application?

Tools that can help

There is no affiliate link on this page: nothing here depends on a paid third-party tool. Our affiliate disclosure explains when we do use them.

Sources checked

• The Authors Guild, AI Best Practices for Authors and AI-Related Model Publishing Contract Clauses.
• The Society of Authors, Artificial intelligence: practical steps for members (advice to seek individual advice and have contracts checked).
• Amazon KDP, Content Guidelines (Artificial intelligence (AI) content).

Start dating your drafts now

Your first Pooled TimeBind is free, with no credit card required. After that, TimeBinds start from US$5, and an ASAP TimeBind is US$15. Click here to start, or see plans and pricing.