Do readers value writing authenticity they cannot detect?

Hwang et al. report a split between writers and readers. Writers who defined authenticity by source saw AI assistance as "a direct threat." Readers in the online survey did not share that concern. The abstract states that "readers could not distinguish AI-assisted work, personalized or not, from writers' solo-written work and showed positive attitudes toward writers experimenting with new technology for creative writing." The excerpt adds only that readers' responses "showed less concern about human-AI co-writing."

The first note's claim suggests why this gap could be structural. If authenticity lives in the process and in the writer's experience, a reader who sees only the finished text has no direct access to what writers say they protect. The reader study, as the excerpt presents it, asks readers to judge written work, so its findings speak to what readers can detect in the text. The excerpt lists the reader-side statistical models, a Wilcoxon test against a midpoint and mixed-effects models with reader and author random effects, but gives no results. The indistinguishability claim therefore rests on the abstract alone.

The contrast with Does AI writing assistance change how readers perceive the writer? is informative. That study reports that AI assistance shifts how readers perceive writers, across 29 dimensions. Hwang et al.'s readers, by contrast, did not register AI involvement. The two findings measure different things, perceived traits against whether AI was involved, so both can hold. Together they sharpen the question: readers may register a persona shift without being able to attribute it to AI, and the excerpt does not say how readers would respond if told AI was involved. Does polished AI output trick audiences into trusting it? offers one candidate link. If audiences read polished surfaces as signs of judgment, finished text could hide process even from careful readers. The excerpt does not test that link.

The excerpt does not establish that readers' lack of concern survives disclosure, whether readers would value a process they cannot see, or whether 30 avid readers stand in for the audiences writers care about. The implication is narrow. The reader result supports the claim that these readers did not detect AI involvement and were not concerned by it. It does not support the claim that process-level authenticity is unimportant to readers. Until a study varies whether readers are told about the process, the question stays open.