Writers did adapt the suggestions they kept — rewording them, adding personal anecdotes, making them sound like their own. But the underlying framing and concepts from the AI survived into the final text, even in passages writers composed themselves.
Writers accepted or rejected suggestions based on whether they agreed with them, regardless of if they had come up with the ideas themselves. They felt in control as they could always reject or edit, but they didn't choose which ideas got surfaced in the first place.
What we saw: writers read AI suggestions before their own ideas had a chance to form. Ideation gave way to reading and evaluating.
As one participant put it: "Well, now I'm not thinking about what I'm writing. I'm reading the sentence."
We analyzed 1,291 co-writing sessions and interviewed 19 participants who wrote an opinion text on a controversial topic. Some wrote with an AI assistant supportive of a position, some with an AI whose suggestions leaned critical, and some wrote on their own, without AI assistance.
Writers with a Positive AI ended up writing about different topics than writers with a Critical AI, and both wrote on a narrower range of topics than writers working alone.
How are AI tools changing our writing process? In our #chi2026 paper, we find a shift in the way writers write and engage with ideas: from generating ideas to reacting to the ones AI surfaces. We call it reactive writing.