AI Videos Starring 'Someone Like You' Cut Health Myths by 30%

Researchers wanted to know why health misinformation sticks. Their theory: false beliefs often survive not because people lack facts, but because those beliefs are tied up with community and family norms. So they built AI-generated videos of a presenter and tested two versions — one that looked and felt like a woman from the viewer's own community, and one generic presenter with no local cues.

The experiment ran in suburban India with 434 women who had limited reading skills, a group often left out of written health campaigns. The results were striking. Women who watched the culturally familiar presenter dropped their belief in false health claims by 30% — nearly twice the drop seen with the generic presenter. And the change stuck: three weeks later, the improvement was still there. In follow-up interviews, women remembered the familiar presenter as a real person they trusted, but described the generic version as just "a video."

But there was a clear limit, and it's a revealing one. The approach worked best on beliefs that communities reinforce — for example, the false idea that a woman is to blame for infertility. It did almost nothing for simple knowledge gaps, like understanding how vaccines work. That tells us misinformation isn't only a missing-information problem. It's a social problem, and it needs a social fix.

THE CATCH: The same technology that builds trust in seconds can be pointed the other way. If a familiar face makes people believe true things more easily, it can make them believe false things too — that's the deepfake risk in one sentence. This was also one study, in one region, with 434 people, so we shouldn't assume it works everywhere. But it's cheap, fast, and easy to copy, which is what makes it both promising and a little worrying.