Is using AI cheating? The six fears I hear and what I actually think

TL;DR: No – using AI isn’t cheating when what you hand off is expression and what you keep is judgment. But the fear deserves better than a pep talk, because it points at something real: outsource the thinking and you outsource the development, and then you’ve cheated yourself no matter what the rules say. I’m Julie Kaiser, I’ve written for regulated biotech and pharma since 2011, I write with AI every day, and here are the six versions of this fear I actually hear, and how I decide when AI belongs in the work.

The six fears I actually hear

The cheating question is never one fear. On the surface it sounds like a rules question – have I deceived somebody, have I passed something off as my own? But it’s more nuanced than that. It comes down to six distinct worries.

Lack of originality

This one’s about pretending: claiming a skill you don’t have, trying to get recognized for a talent that isn’t yours. Pretending is worth being afraid of.

Skill atrophy

I’ve seen people say in forums that they haven’t written anything themselves for years – not since ChatGPT came along. And I catch myself asking for help composing a stupid email. So this part is real. We get lazy.

Theft from training data

Did the machine learn its craft on work it had no right to? The training-data question is real and unresolved. I won’t pretend it’s settled, because it isn’t – it’s a complicated issue, and it’s a legitimate part of the argument.

No copyright, no ownership

There’s a claim you’ll run into again and again: that under US law, AI can’t hold copyright, so it has no rights to hand you for the content it generates. I don’t know how solid that is – the legal picture is unclear and I’m not a lawyer. But the ownership worry underneath it is real.

Dishonesty without disclosure

Passing something off as your own without telling people. Disclosure gets its own section below, because “it depends” deserves to be shown, not asserted.

Cheating yourself out of development

Skip the thinking and you skip the development. I think this one lands hardest for people who aren’t doing this for a living: I do this as my job – I’m not writing to develop, so I’m not cheating myself out of anything. But for somebody who’s supposed to be learning and growing through the work, I can see exactly where this fear comes from.

What cheating actually means (and doesn’t)

Cheating needs a rule, a contest, and a deception – you broke an agreement about how the work would be done, in a setting where that agreement was the point. School is that setting, which is why the question is genuinely hard there.

Most working life isn’t. Nobody hired you under an agreement that you’d type every word unassisted; they hired you for an outcome and the judgment behind it. So for professionals, “did I break a rule” is rarely the honest question. The honest questions are the two hiding inside the six fears: did I deceive anyone, and did I shortchange myself? The rest of this post works through both – and the output-trust version of the whole question, whether you can believe what AI hands you, is the flagship hallucination post.

How I decide whether to use AI

I’d love to hand you a single rule, but what I do depends on why I’m writing and who it’s for.

This blog is heavily AI-generated. I’m vibe writing it, and I see no problem with that: these are topics where I hold the information and the perspective – my opinions, my stories – and I disclose how the posts get made. The value of getting real, useful content out quickly beats me toiling over every word by hand. At least I hope so – the bet is that what I’m making actually helps people.

Client work is grayer. A client pays me with the expectation of my expertise, and as the AI tools get better at writing, it’s tempting to outsource more of the work to them. For certain clients who want me to work that way, I do – and it’s a conversation I expect to have more often as the capabilities improve. Honestly, I don’t think my clients want to pay me to produce every sentence by hand, untouched by any tool. In the fields my clients work in, people are busy and underfunded, and the work has to fit inside real budgets. If I can deliver the scientific judgment, the nuance, and the quality of writing for less money, that’s worth doing – and it’s probably the direction we’re heading.

Should I disclose I used AI?

It depends on what the person on the other end thinks they’re getting – and I mean that as a test, not a dodge. If a client pays for your judgment and you use AI to shape the expression, that sits inside what “hire a professional” has always meant. If a reader follows you because they think they’re getting your thinking, and it’s actually AI’s thinking with your name on it, that’s a disclosure problem. And if someone follows you because they want your thinking, and it is absolutely your thinking with AI’s typing? No problem at all – though I still think disclosing is a good idea.

The test: did I give the buyer what they thought they were buying?

If you’re worried you’re cheating

The worry is diagnostic, and I think it’s mostly good news: the people hollowing out their own skills don’t tend to ask this question. If you’re actually asking it, you’re probably in a good position. So run two checks.

Could you defend every claim in the work without the AI in the room? And would the person receiving it feel deceived if they watched you make it? Pass both and use the tools without the guilt. Fail either and you don’t have an AI problem – you have a specific fix to make, and now you know which one.

The mirror of this worry, being falsely accused by a detector when you wrote it yourself, is its own post: AI writing detectors and false positives.

The rule underneath all of this

Use the machine, and keep the judgment. What you sign should be yours – that’s the entire ethics of it, as far as I’ve worked it out, and none of us has the final answer yet.

The place where this rule becomes practice, the editorial gate everything of mine passes through, is in How I built an AI that actually knows my work.