When people edit an AI draft, does the time they spend reveal real judgment, or do most just send it along?

This explores whether the minutes someone spends editing an AI draft are a visible sign of human judgment, and what that sign can and can't tell us about who is accountable for the result. The clearest evidence comes from freelance job applications. When the same worker spent more time editing an AI-written cover letter, they were more likely to be hired, even though most workers sent drafts with almost no changes Does editing time on AI drafts predict hiring success?. That result is correlational, not proof that editing causes hiring. It still suggests that clients notice the difference between a draft someone has worked through and one they forwarded. The default behavior is the bigger story. In a separate study, writers edited AI paragraphs only 23% of the time, and the edited versions stayed 96% similar to the original. The AI's opinions and slants therefore reached readers almost untouched Do writers actually edit AI-generated text before publishing?.

Why would editing matter so much? One argument is that once AI makes first-pass thinking cheap, the scarce thing is a person who checks the output, puts their name on it, and answers for it What makes accountable judgment scarce when AI cognition is cheap?. Seen that way, editing time is one of the few traces of accountability you can measure. It also connects to how people feel about the text. Writers feel more ownership of AI-generated text when they have more influence over it. Personalizing the AI model doesn't produce the same effect Does user control over AI text shape feelings of ownership?. Editing is how a draft becomes yours, both to you and to the person reading it.

Time is a weak measure on its own, though. AI doesn't reliably shorten tasks. It moves time from doing the work toward writing prompts and figuring out what the model produced Does AI really save time, or just change how we spend it?. So a long session could mean careful judgment, or it could mean confusion. A study of 640 employees found that reviewers caught more AI errors only when the reasoning they needed to check the output was easy to recall at review time. Explaining things in their own words or getting retrieval cues helped Can reviewers access what they know when checking LLM outputs?. Someone can spend a long time staring at a draft and still miss what matters if that knowledge isn't available to them in the moment.

There's also a social twist. People expect to be seen as less competent and less diligent when they use AI, so they tend not to tell their managers Do people fear judgment when they use AI at work?. That pushes editing out of sight at exactly the point where it could show diligence. Collaborative writing tools point the other way. Writers working in pairs wanted to see when and where their partner used AI, partly so they could check the AI-written parts, although some found full visibility uncomfortable Do writers want to see each other's AI prompts in shared editors?. One more risk runs underneath all of this. People can come to treat AI output as evidence of their own ability, a self-perception error separate from hallucination or over-reliance on automation How does AI-assisted work reshape how people see their own abilities?. Editing helps here because it makes the line between the person's contribution and the machine's harder to blur. The takeaway is that editing time works less as a productivity number and more as a rough, partly hidden record of whether anyone actually took responsibility for the words.