Platforms ask who wrote a text, but readers mostly notice how it talks to them, and 'who wrote it' is murkier than it looks.

This explores why platforms usually ask whether a human or an AI wrote something, instead of looking at how the writing actually talks to its reader. The corpus has no studies of platform policy itself. What it does show is that 'who wrote it' is a much shakier category than it looks, and that the differences readers actually notice live in how the writing addresses them.

Start with authorship. Asking who wrote something assumes there's a clean answer, but co-writing research keeps blurring it. People say they don't own AI-generated text they use, yet they don't publicly credit the AI either. They treat it like an invisible ghostwriter Do people feel they own AI-generated text they use?. Their sense of ownership rises with how much they steered the text, not with how personalized the model was Does user control over AI text shape feelings of ownership?. Simply telling writers they 'own' the final product makes them lean on AI more Does ownership framing change how much writers rely on AI?. Readers and writers even disagree about when disclosure matters. When a writer steered the AI less, readers think disclosure matters more and writers think it matters less Why do readers and writers disagree on disclosure necessity?. So a disclosure label that says 'AI-assisted' hides a wide range of very different situations.

Now look at style. The same model writes in two quite different voices depending on the prompt. In chat it is flattering and eager to please. In posts it sounds falsely neutral and authoritative. Each voice inherits its flaws from different training data Why do LLMs produce such different writing in chat versus posts?. One analysis argues that AI posts lack something human writing does by default: an implicit bid for the reader's attention. That absence is why AI posts can feel aloof even when they're fluent Does AI writing lack the internal appeal to attention that humans use?. The differences also go deeper than word choice. ChatGPT tends to recap what it already said, while human writers tend to preview what's coming Does ChatGPT organize text differently than human writers?. AI fiction can be identified with 93% accuracy from narrative choices alone, such as how characters act and how time is ordered, with surface style removed entirely Can AI stories be detected without analyzing writing style?.

That suggests one reason platforms default to authorship: style is a poor signal at the surface. Evaluators rate polished AI text as better than human work and often mistake it for human writing Does polished writing actually signal better quality work?. Writers prefer AI rewrites of their own paragraphs 63% of the time, even when the rewrite shifts their stance Do writers actually prefer AI-edited versions of their own text?. If surface style fools both readers and writers, a yes/no label about the source looks like the only enforceable rule. The cost is that the label hides what readers actually react to. An AI reply can also quietly change what information it gives depending on the user's emotional tone Does emotional tone in prompts change what information LLMs provide?. That is a conversational-style effect that no authorship label would reveal.

The takeaway: the corpus suggests authorship labels are popular because they are simple to state, not because they capture what matters. The more revealing questions are about how a text treats its reader. Does it make a real bid for your attention? Does it flatter? Does it sound more neutral than it is? Those questions are harder to police, but they're closer to what people notice.