Should You Tell People You Used AI at Work?

You used AI to draft the report. Do you say so? The instinct is that honesty is always right, but the honest answer here is more uncomfortable: research suggests that disclosing AI use can lower how much colleagues trust your work, even when the work is good. And saying nothing carries its own risk, because being found out later is far worse than mentioning it upfront.

So the useful question is not whether to be honest. It is where the line sits between a tool you do not narrate and a contribution someone deserves to know about. Here is a practical way to decide.

Why disclosure feels risky (and sometimes is)

Work by Atlassian's team and academic researchers has found a real trust penalty attached to AI disclosure: people who reveal they used AI are sometimes judged as less competent than people who quietly did the same thing. That finding is uncomfortable, and it is also why blanket advice fails. If you treat it as "never mention it", you eventually get caught in something that mattered. If you treat it as "announce it every time", you tax yourself for using a spellchecker.

The way out is to stop thinking about AI as one thing. Nobody discloses that they used spellcheck, a calculator, or a template. People do expect to know when the substance of the work was not yours.

The spectrum: from no need to always say

No need to mention. AI fixed your grammar, tidied your phrasing, summarized a document for your own reading, helped you brainstorm before you decided, or translated something you then checked. The thinking and the judgment stayed yours. This covers most of what our office prompt library is used for.

Worth mentioning casually. AI produced a substantial first draft that you then edited, or did the initial analysis you verified. A single line does it: "I used AI for a first pass and then reworked it." Said lightly, it reads as competent process, not confession.

Always disclose. Anything where the reader assumes personal effort or personal opinion: performance feedback about a person, a reference letter, condolences or congratulations, creative work presented as yours, client deliverables where AI use may be contractually restricted, and anything your company's policy names. Also disclose when AI generated facts or figures that others will rely on.

Check the policy before you decide anything

Many companies now have an AI policy, and a growing number require internal disclosure for certain work. That document beats every judgment call in this article, and reading it takes five minutes. If your company has none yet, the spectrum above is a safe default that will survive whatever policy arrives later. The data-handling rules in our guide to using ChatGPT safely at work apply regardless of what you disclose.

How to say it without undermining yourself

Tone does most of the work. "Sorry, I just used ChatGPT for this" invites doubt. "I drafted this with AI and then verified the numbers and rewrote the analysis section" invites confidence, because it tells people you stayed in charge. Emphasize the part you did:

Help me write one short sentence telling my [manager / client] that I used AI on this piece of work. Context: I used it for [what exactly], and I personally did [what you did: verified, rewrote, decided, checked sources]. Confident and matter-of-fact, no apology, one sentence, no jargon.

The part that actually protects you

Whatever you disclose, you own the output. If a number is wrong, "the AI said so" is not a defense anyone accepts, which is why the verification habit in our fact-checking guide matters more than the disclosure question itself. And if a draft does not sound like you, that gap is what makes people suspicious in the first place, so keeping your own voice is quietly part of the answer too.

A simple test before you send anything: could you explain every claim, defend every number, and rewrite any paragraph from scratch if asked? If yes, the work is genuinely yours and disclosure is a courtesy. If no, you have a bigger problem than etiquette.

Where this is heading

Norms are shifting fast. In two years, saying you used AI on a routine draft will sound like saying you used a keyboard, and the disclosure conversation will narrow to the places it belongs: work about people, work presented as personal, and work someone is paying for. Until then, use the spectrum, read your policy, and lead with what you did rather than what the tool did.

If you want to be the person whose AI-assisted work nobody questions, the fundamentals matter most. Our free course ChatGPT from Zero: Your First Week with AI covers them in six short lessons.