A new study says your AI-assisted work only feels like yours if you actually do something with it.
Researchers ran a qualitative survey asking participants to describe two recent tasks they did with AI help: one that felt like their own and one that didn't. The pattern that separated the two wasn't how much AI contributed, but how the person worked with it. People who just approved AI's suggestions tended to disown the result, while those who led the process, iterated, or rewrote the output kept a sense of authorship. Ownership also showed up in unexpected places: participants reported high ownership on tasks they could not have finished without AI at all, as long as they owned the vision even if AI handled the execution. Losing personal voice in the output, or not fully understanding what the AI produced, both eroded that sense of ownership.
The more interesting finding is about disclosure. Whether people admitted to using AI had little to do with how proud they felt of the work. It was shaped instead by community norms and a fear that admitting AI involvement would erase their credit. That's a useful data point for any workplace drafting AI disclosure rules: people aren't hiding AI use because they're embarrassed by the output, they're hiding it because they expect to be penalized for saying so.
It's a small, self-selected sample and won't settle any argument about what counts as "real" authorship. But it does put a name to the specific discomfort of shipping something you technically made but didn't really touch.