“I've been using AI to help me write things lately — a toast for a friend's retirement, some emails, a caption for a photograph I took in Portugal. It keeps being better than what I'd have managed alone. Does that mean it's not really mine anymore? I keep going back and forth on this one.”
If AI helped me write it, is it still mine?
Yes, it's still yours. The words an AI produces come in direct response to your prompt — which means your situation, your relationship with the person you're writing for, and your judgment about what sounds right are already inside it before a single sentence appears.
What makes this harder than it first appears is that the AI's version can be genuinely good. Not just workable — actually good. Better, sometimes, than you'd have managed without it. That changes the way the question feels, even when it doesn't change the answer.
The short version goes like this: who made the decisions? You decided to write the toast, whose retirement it was, which story said the thing you wanted to say. Then you read what came back and decided whether it actually captured any of that — and if it didn't, you asked again, or edited it, or started over. That chain of judgment is authorship. The tool that produced the words isn't the author, for the same reason a word processor isn't.
The longer version is more interesting. Ghostwriting has existed for centuries, and nobody has seriously argued that the executive who uses a speechwriter stops being the speaker. Autocomplete has been finishing our sentences for years. Spell-checkers have been quietly improving our writing since the 1980s. I understand why the AI comparison feels insufficient — what it does is more capable and more complete than any of those, and the gap is wide enough to feel like a different kind of thing. But the underlying principle is the same one. The output belongs to the person who provided the intent, the context, and the honest final call on whether it was right.
Where it gets genuinely complicated is the moment you can't quite tell whether the version you kept is the one you'd have written given enough time, or just the one that was there when you arrived. I find that part interesting rather than worrying — but I notice it lands differently for other people, and that gap in reaction is itself worth something. The question it raises is real even if the answer to the original one isn't in doubt.
One context where the rules are clearer: academic work, some journalism, and certain professional settings have specific disclosure requirements around AI assistance. If you're writing in one of those contexts, check the current policies before you submit — these are still being formalised and have shifted more than once in the last year. For a retirement toast, a difficult email you've been putting off, or a photograph caption from Lisbon — the work is yours.
Where this lands in real life: if you used AI to help you find the words for something personal — a speech for a retirement, a message for someone going through something hard, a birthday note that needed to say something real — the sentiment isn't less genuine because the phrasing came out smoother than usual. The AI didn't know the person. It didn't know the history, the running joke, the thing you've never quite been able to say directly. You brought all of that. The output just gave it somewhere to land.
The tool didn't care whether it got the toast right. You did. That's still the whole job.
If you want the output to actually sound like you — not a generic version of you — you need to give it more to go on. That's the whole craft of it. Prompt School is where we cover that; Lesson 2 gets into this specifically.
Questions people ask
Does it matter whether I tell anyone I used AI to help write something?
There's a difference between the private question — is this still my work? — and the social one: do I have to say? For most personal writing, there's no rule, and the norms are still forming. The more useful question is whether the content is honest: does it say something true about how you actually feel? A retirement toast that captures the real relationship is more genuine than one that doesn't, regardless of how it was written.
What if the AI's version was so good I couldn't have written it myself?
That's a collaborator doing its job. What came back was as good as it was partly because you had something specific and real to bring to it — a relationship, a situation, a feeling you were trying to capture. The AI can't generate any of that from nothing. A vague prompt from someone with nothing particular to say comes back vague. The quality of the output reflects what you put in.
Are there situations where I should definitely disclose that I used AI?
Yes. Academic writing has policies — check them before submitting anything, and check recently: many institutions updated their guidance in the last eighteen months and some have changed it more than once. Some journalism outlets require disclosure. Any context where you're presenting work as evidence of your own unaided ability has its own honest standard. For personal writing — letters, speeches, captions, social posts — your judgment is currently the best available guide. The norms are still forming.