If clinicians already rate advice higher when they think an expert wrote it, what happens when they know who actually did?

This explores whether telling clinicians who wrote a piece of medical advice, before they rate it, would change their scores. The corpus points to an unexpected answer: their ratings already depend on authorship. What they depend on is who the clinicians think wrote the text, not who actually did. In one study, clinicians preferred advice they believed came from an expert 93.55% of the time, even though their guesses about the author were no better than a coin flip. Their quality and empathy scores followed that guess, not the text's real source Does the label on advice shape how clinicians judge it?. So showing authorship up front wouldn't add a bias that isn't already there. It would point the existing bias at the right target. Right now, good AI advice that gets mistaken for expert writing receives the 'expert' boost. Once labeled, that same advice would probably lose it.

Which way the scores would move depends on what kind of writing is being judged. In a study of 261 readers, revealing AI authorship lowered perceived trust, caring and likability, and the drop was steepest for interpersonal writing. Readers felt AI couldn't really empathize, so using it for personal messages seemed like a social violation How does revealing AI authorship change reader trust?. Patient advice is largely interpersonal, and empathy was one of the scores the clinicians shifted. That suggests the empathy rating, more than the accuracy rating, would fall hardest once an AI label appeared. Another study found the opposite pattern for scientific abstracts: disclosing authorship raised trust and quality ratings across the board, apparently because readers rewarded the transparency itself Do reader judgments reflect actual authorship or just their beliefs?. Disclosure isn't automatically a penalty. In factual, technical writing it can come across as honesty. In caring, personal writing it tends to come across as a lack of care.

The size of this effect is easier to see in studies that controlled for the text. When identical literary passages were labeled 'human,' human judges rated them 13.7 points higher Do authorship labels bias how we judge literary quality?. Labels also change how strictly people apply the rules. AI judges excused a broken writing constraint when told a human wrote the passage, while human judges became stricter in that case Do authorship labels change how AI judges evaluate rule violations?. Without labels, evaluators fall back on surface cues: polished AI text is often mistaken for human work and rated higher Does polished writing actually signal better quality work?, and the number of citations can raise trust even when the citations are irrelevant Do users trust citations more when there are simply more of them?. Taking the label away doesn't produce a neutral judge. It hands the decision to a different shortcut.

One point that matters for medical evaluation: the label bias was about 2.5 times stronger in AI evaluators than in humans Do authorship labels bias how we judge literary quality?. If hospitals start using LLMs to grade clinical advice in place of clinicians, showing authorship could skew the scores even more. The corpus has no study that directly compares clinicians rating labeled and unlabeled advice side by side. The evidence above strongly predicts that labeled ratings would differ, with empathy likely the most affected, but that specific experiment hasn't been run here.