should you disclose that you used ai?
the anxious version of this question is "will i get caught." the useful version is "would i sign it."
disclose ai's role wherever it would change how someone judges or relies on the work: authorship claims, advice they will act on, anything with legal, medical, or financial weight, and anywhere a client or employer has a stated policy. you do not need to annotate every spell-check. the honest test: would hiding it feel like a problem if it came out? this is the skill anthropic calls diligence, the fourth AI fluency skill: using ai responsibly.
diligence has three parts
- creation diligence. choosing the right tool for the job and using it thoughtfully. a sensitive client file does not go into a random free tool with unclear data terms.
- transparency diligence. being honest about ai's role with the people who need to know. not a confession, a norm.
- deployment diligence. verifying and vouching for whatever you use or share. you checked it, so you can stand behind it.
when to disclose, and when it's just noise
disclosure is about what the other person is relying on, not about ai being present. here is the practical line.
| disclose | no disclosure needed |
|---|---|
| ai wrote content presented as your original authorship | ai helped you brainstorm or outline your own work |
| advice they will act on (legal, medical, financial) | ai fixed your grammar or reformatted a doc |
| a client or employer has a stated ai policy | ai sped up research you then verified and wrote up |
| ai stood in for a human they think they're getting | routine internal drafts you fully own and edit |
the rule that settles it: would you sign it?
accountability does not transfer to the model. whatever you send carries your name, not the ai's. so the whole ethics of it compresses to one question: would you sign it, and defend it in a room? if yes, it ships. if no, it does not. that keeps you honest without turning every email into a disclosure form. it pairs with discernment: you can only vouch for what you actually checked.
why this is a trust advantage, not a burden
most businesses are quietly nervous about ai. the person who can say "yes i use ai, here is exactly where, here is what i still own, here is how i check it" is the person they trust. transparency is not the tax on using ai. it is the thing that makes clients comfortable letting you use it. i got certified through anthropic partly so i could answer that question with receipts.
the takeaway
you can use ai for real work. pick the tool with care, disclose where it changes what someone relies on, and own what you ship. would you sign it is the whole test.
want ai set up so you can always vouch for it?
the free system builder sets ai up on your work with a human on the send and your standards baked in, so what ships is always something you'd sign. built in claude, today.
get the free system builderor just follow along. new field notes most weeks on x, instagram, and tiktok.