← All articles

Writing in your name with AI: where help ends and impersonation begins

Picture a self-employed physiotherapist. She answers twenty patient messages a day, and every night she carries three she cannot work out how to answer without sounding cold. One day she pastes them into an AI assistant and what comes back is better than what she would have written. In time, she asks it to answer her Google reviews too. Then to write one — a grateful one, "from a happy patient" — to encourage others. No single step is much bigger than the one before it.

(The case is illustrative. It is not a real person.)

Where did she stop being herself?

Why this is not obvious

Because other people have been writing things for us for decades. A secretary drafted the letter. An accountant drafts the filing to the tax office. A friend fixes your cover letter. Nobody calls any of that a lie. Doing it with AI feels like the same thing, only cheaper.

The law does not draw the line here either, at least not yet. The transparency duties in Article 50 of the EU AI Act have applied since 2 August 2026, but they look mainly at whoever builds the systems — who must mark generated content — and at specific cases such as deep fakes and chatbots talking to the public. An email of yours, drafted with help from a tool, is not in scope. I went through what it does cover in the piece on Article 50.

The best argument from the other side

It is this: what matters is the result. If the message is clear, kind and true, who cares who typed it? No client has a right to know which text editor you use. And for a lot of people who write badly, or who are writing in a language that is not theirs, AI is simply fairness — it lets them sound as polished as someone who had more schooling.

I have little to object to there, and I use AI to tidy up my own prose.

What I think

That the question is not who wrote it. The question is what the text claims about the world and about you.

An email drafted with AI says: here is my answer, and I stand behind it. It is still yours, because you read it, corrected it, sent it and put your face to it. The AI acted as an editor.

A fabricated review says: I lived through this. And that is not true. Here the AI is not helping you express something of yours — it invents an experience and attributes it to someone. That is not help, it is impersonation. European consumer law already treats adding fake or manipulated reviews, or commissioning them from a third party, as an unfair commercial practice, whether a person or a machine wrote them.

Between the two extremes there is a grey zone, and there I fall back on a simple test: if the person reading it knew exactly how it was written, would they feel deceived? A polished email, no. An emotional apology written by a machine and sent as though it were a personal gesture, possibly. A review from a customer who does not exist, obviously.

What worries me is not the clear frauds, which are easy to condemn. It is the gestures that count precisely because they cost something: a letter of condolence, an apology, a thank-you. Delegate those entirely and the person receiving them reads correct words and loses the thing they actually cared about, which was knowing that somebody spent time on it.

What this means if you run something small

And if you run a platform rather than a small business

The same line, moved up a level. The question stops being "did a person write this" and becomes "does this account claim to be someone it is not" — which is where impersonation policy, review integrity and inauthentic-behaviour detection all live. The test survives the change of scale: not who produced the text, but what it asserts about the world and about the person it is attributed to.

Sources

This article is general information, not legal advice. For a specific decision with legal consequences, consult a qualified lawyer.

About this article

iaetica.cat is a space for writing in the open: what I am learning about artificial intelligence, its ethics and its law.

— Laura Graells