Personal Perspective: AI Is separating two things we never had to tell apart. #
Posted July 30, 2026 [ Reviewed by Kaja Perina
](/us/docs/editorial-process)
Key points
- Writing used to mean the thinker and the writer were the same person.
- AI can now write fluent sentences with no lived experience.
- The question isn't who gets credit, but if we can tell if there's a mind behind it.
The author.
Looking back, that word carried a single and powerful claim. The author was the one who had the thought and wrote it down. It was the same person and the act without a gap between them. This is what gave writing its authority. It wasn't just the elegance of the sentence, but the assumption that a mind stood behind it. Importantly, being a writer was inexorably connected to being a thinker.
AI may be the first technology with the power to separate the two.
Start with what we can see. Linguistic authorship is the words themselves. It's the structure and phrasing that carries an idea off from one mind to another. You can point to a well-crafted sentence and admire it for only its construction. And for a long time, that visibility was enough, because the words were the only evidence we had of the human mind that produced them.
Good sentences implied good thinking. We had no reason to doubt this inference.
Cognitive authorship is different, and much harder to see. It's the judgment that decided which idea mattered combined with the reasoning that shaped it. None of this appears on the page. We trusted that it had happened, because the words in front of us seemed to require, if not demand, it.
Today, an LLM can produce linguistic authorship that is both competent and elegant. But I'll argue that there's no cognitive authorship behind it at all. Yes, there's a statistical power that shapes the sentences and mimics thought. But there's no resistance or personal cost—just output. And yet the sentence seems like someone agonized over every word. This is the part that we, you and I, need to think about.
Textual fluency stopped being evidence of anything.
I say this as someone who uses these LLMs daily. I bring formed and half-baked ideas to a model and ask it to test them in iterative dialogue. That's not nothing, but I don't think it's quite writing either, at least not in the traditional sense where the sentence and the thought arrived together. I notice the difference. I'm not always sure the reader really can.
Maybe that's the real question the diagram is forcing us to confront. It's not who gets credit for the words on the page, but whether we can still tell when a mind was actually behind it. In the past, we didn't have to ask. But today, we may not get to avoid it much longer.