I’m not the only one wondering. In a survey conducted in March 2026, Lean In found that 29% of women worried they'd be perceived as cheating when they used AI, compared with 22% of men. Not sure about you, but for me, March 2026 seems a lifetime ago in AI years.
I don’t know what that percentage would be today. I do know that the question doesn’t seem to be going away. Neither do the questions about whether men and women are using AI differently and what happens when they do.
But back to cheating.
When does AI cross the line from helping you do the work to doing the work for you?
I know “it depends” isn’t the most satisfying answer to that question, but I think it’s the right one. After all, doesn’t the line move depending on what you’re being asked to do?
Use it to help you prep for an interview? Probably fine. Use it to tell you what to say during the interview? That feels pretty different to me.
Use it to help you figure out why your Excel formula isn’t working. I would think that’s a great use. Ask it to do the analysis and write the summary and then hand it to your manager without even reading, much less understanding, what it did? Feels wrong.
Use it to help you turn your notes into a first draft? Maybe. Ask it to write something on a topic you don’t understand and then put it out in the world under your name. Seems like a clear no, right?
I don’t think you can answer whether AI is cheating unless you start with what you were supposed to be doing in the first place.
If your manager asks you to pull together key themes from 50 pages of customer feedback, is the point really to prove that you can read 50 pages of comments and sort each one into a bucket? My guess is that the point is to come back with something useful. AI could help here.
But what if the next ask from your manager is what you think the feedback means? Or how your team should act on it? Or what your company should do about it? Or whether it's even worth paying attention to in the first place?
Isn’t that judgment, the very thing employers keep telling us they want?
Maybe that “line” from earlier isn’t whether or not using AI helped. Maybe it’s whether you let it do the part you were actually being asked to contribute. Of course, that assumes you know what that part is.
Your manager may care deeply about your analysis. Do the numbers work? Were you working with the right data? They may care much less about who wrote the first draft of the memo. It would be great if assignments came labeled: THIS IS THE PART WE’RE EVALUATING. They don’t.
So when the rules and the expectations aren’t clear, what do you do?
For me, the answer has something to do with responsibility.
There’s a phrase in a STIHL tools commercial that keeps echoing in my mind: “Doing it yourself should never mean doing it alone.” What a great way to describe responsibility, right? You don’t have to do every little part of the work yourself to be responsible for it. Haven’t we been using tools, templates, other people’s insights and feedback forever?
If AI helped you analyze the data, can you tell whether the analysis makes sense? If it found a pattern, did you check whether the pattern was actually there? If it gave you a statistic, did you check the source? If it wrote the first draft, did you read and refine it before turning it over?
And if your manager asks, “Why did you recommend this direction?” can you answer without going back to AI?
So here’s one test: could you explain how you used AI if someone asked? That is, could you explain what you handed to AI, what it handed back to you, and what you did from there?
That seems like it should be a fair test. Except there’s a wrinkle.
People may judge you differently when they know you used AI.
Researchers at Duke ran a series of experiments involving more than 4,400 people and found that employees who used AI were perceived as less competent, less diligent and lazier than employees who didn't. And the people who imagined using AI expected that judgement. They said they’d be less likely to tell their managers.
And then there’s the gender piece.
In another study, more than 1,000 software engineers evaluated the exact same piece of code. Some were told that the person who produced it had used AI. Some weren’t. The work didn’t change. Their perceptions of the person who did it did. And the competence penalty for using AI was larger when the engineer was a woman.
Which leads us back to Lean In.
Maybe women who worry they’ll be seen as cheating when they use AI have a reason to worry. There may be an actual cost to saying they used it.
So where does that leave us?
I think you have to start with yourself. Before you worry about whether someone else will think you cheated, make sure you know that you didn’t.
Know what you were being asked to contribute. Know what parts of it you asked AI to do for you. Check what it gave you back. And make sure that when the work goes out with your name on it, you can explain it, defend it, and take responsibility for it.
True, that still doesn’t tell you exactly where the line is every time.
I don’t think there is one. At least not one universal line. As AI gets better at doing more of the work, we’re going to have to get better at deciding which part of the work should still be ours.