Tell readers an AI wrote a passage and they'll rate the identical words lower, even though nothing on the page changed.

This explores how much the label "AI was used here" changes how readers judge a text, apart from what the text actually says. The corpus's short answer is that the label often matters more than the words. In one study, people rated identical literary passages 13.7 percentage points higher when they were told a human wrote them. AI models asked to do the same judging were about two and a half times more biased, at 34.3 points Do authorship labels bias how we judge literary quality?. So if you hoped AI judges would be the neutral referee, they turn out to react to who wrote something even more than people do.

The penalty isn't the same everywhere. When readers learn AI was involved, they lose trust in the writer and see them as less caring and less likable. The drop is steepest in personal, relationship-focused writing, because readers treat AI as unable to feel real empathy and see its use there as breaking a social expectation How does revealing AI authorship change reader trust?. Readers with more AI literacy show smaller drops, and some even view AI use positively, so familiarity seems to soften the reaction Does AI literacy reduce the damage from AI disclosure?. Readers and writers also disagree about when disclosure is owed. Readers consistently want it more than writers think they need to give it, especially when AI text was pasted in directly. How much effort the writer put in made no difference to those judgments Do readers and writers differ on AI disclosure necessity?.

Here's the twist: without the label, readers mostly can't tell. A review of 30 studies found that people spot AI content at about chance level across text, images and voice Can people reliably spot content made by AI?. AI text does differ measurably from human text, for example in vocabulary variety and how evenly words are spread. But even trained linguists miss these differences, and newer models drift further from human patterns while getting harder to spot Can humans detect AI text if machines can measure it? Can human judges detect measurable differences in AI text?. Because readers can't see the difference, suspicion fills the gap. Comments accused of being AI-written show no features that actually separate AI from human writing. That suggests accusations act more like gatekeeping than detection, and the people they wrong are human writers Do unfounded AI accusations harm human writers instead?.

The part readers rarely think about is that not knowing doesn't mean judging neutrally. In a study of nearly 3,000 writers and 11,000 readers, AI assistance shifted how readers saw the writer on all 29 traits measured, making them seem more confident, more extreme, more agreeable and more privileged Does AI writing assistance change how readers perceive the writer?. Writers edited AI-suggested paragraphs only 23% of the time, and their edits left the text about 96% the same, so that shifted voice reaches readers almost untouched Do writers actually edit AI-generated text before publishing?. One theoretical line argues this is expected: readers interpret AI text with the same habits they use for human text, so it has the same social effects Does AI text affect readers the same way human text does?. Another notes that we have learned to automatically discount advertising, but we don't yet have a built-in skepticism for AI-generated text How do we learn to read AI-generated text critically?.

Put together, knowing about AI use changes judgments a lot. But the bigger effect may come when readers don't know: the AI's shaping of how the writer comes across passes through unnoticed. Disclosure brings in bias against AI, and silence lets the AI's influence pass as the writer's own voice.