Platforms such as Meta now label content made with AI. Does a label like that protect the reader from a political message?
First, a guess. What will an "AI-generated" label change?
The same political message, word for word"Expanding the programme would cut waiting times for thousands of families." AI-generated
How far the message moves policy support
1,601 Americans, then 3,861 in a probability sample
no changeSupport moved 9.74 points on a 0 to 100 scale. With the label, the difference was 0.11. The study could see a gap of about 3 points, nothing smaller, and a second study agreed.
No label
AI label
What people think of the content
accuracy, trust, willingness to read
small dropJudged accuracy fell 2.66 points (a label saying the item was false moved it 9.33), with less trust and less willingness to read. Small to moderate.
Belief in a flagged, misleading claim
7,579 Americans, real posts with misleading AI images
real dropA label of any kind lowered how much people believed the post's central claims. This is where labels work.
Figures from the article's sources: Gallegos et al. (2026); Wang, Sturgis and de Kadt (2026); Altay and Gilardi (2024); Toff and Simon (2025); Zoizner et al. (2025); Wittenberg et al. (2025). The message shown is invented; the studies used their own.