Disclosure: telling people you used it

The question is coming

Clients ask. Students are asked. Editors, funders, courts, universities and procurement teams have started to ask. It is worth deciding your position before somebody asks it under pressure, because an improvised answer tends to be either defensive or dishonest.

The principle

Disclosure is owed when the reader is relying on your personal judgement or authorship as such. It is not owed for every tool you used to produce a document.

Nobody expects to be told that a spreadsheet did the arithmetic, that spell-check caught a typo, or that a template supplied the structure. Those are production tools. The reader's reliance is on your conclusions, and the conclusions are yours.

The line moves when one of three things is true:

1. Authorship itself is the thing being assessed. A student's essay, an exam, a writing sample, an application testing your ability to write. Here the process is the product and undisclosed use is straightforwardly dishonest.
2. A rule requires it. Many journals, courts, universities, funders and public bodies now have explicit policies. Several courts require a certification that filings were checked. Read your own institution's rule; do not infer it.
3. The reader would feel misled to find out. The most reliable test, and the one to fall back on when the other two are silent. A condolence letter. A personal reference. A therapist's reflective note. A message that traded on being written by you, personally, for this person. The offence there is not procedural, and no policy will get you out of it.

The rules that already exist, and where to find them

Before reasoning from principle, check whether somebody has already decided for you. Rules now exist in more places than most people realise, and they are specific rather than general.

• Courts. Several jurisdictions require a signed certification that any AI-assisted filing was checked by a human, and some require disclosure of use. Practice directions are published; read your own court's.
• Journals and publishers. Most major publishers now require that AI use be described in the methods or acknowledgements, and almost all refuse to list a model as an author, on the ground that an author must be able to take responsibility.
• Universities. Policy is usually per-assessment rather than institution-wide, which is why students get this wrong. The rule for the coursework is in the coursework brief.
• Public bodies and procurement. Increasingly written into tender terms and framework agreements. If you sign one, that clause governs your delivery, whatever your internal policy says.
• Your own engagement letters. Look at what you have already promised clients. Some standard terms now say more about this than firms realise.

Ten minutes reading the rule that actually binds you is worth more than any general position, including this lesson's.

Practical positions

Client work. Most professional services are sold on judgement and outcome, not on keystrokes. You do not owe a running list of tools. But if a client asks directly, answer directly — and if your engagement letter or their procurement terms say anything about AI, that governs. An increasing number do. Read them before signing rather than after delivering.

Internal work. Say so casually and often. "I drafted this with Copilot and checked the figures against the ledger" costs nothing, normalises the practice, and tells colleagues what has been verified and by whom. Teams that talk about it openly make far fewer errors than teams where everyone is quietly doing it.

Public-facing content. If it is presented as a person's writing, and it was substantially generated, say so. Generated images in an otherwise factual piece should always be labelled, because the alternative is asking readers to guess which parts of your output are real.

Teaching and assessment. If you are a teacher: state the rule for each task, and state the reason. "You may use it to plan and to check; the final analysis must be yours" is a policy students can follow. "No AI" is a policy nobody can follow and nobody can enforce, and it mainly teaches concealment.

What you never say

"The AI did it."

This has failed as a defence in every professional setting where it has been tried — courts, regulators, employers, editorial boards. The signature is yours. The judgement is yours. Disclosure tells people how the work was made; it does not transfer responsibility for it, and attempting to use it that way converts a manageable error into a credibility problem.

The sentence to have ready

Something like: I use AI tools to draft and to check my work. Everything I send you, I have verified and I stand behind.

That answers the question, describes a defensible process, and puts the accountability exactly where it belongs. It is also true, provided you have done the checking — which is what the last module of this course is about.