Meta-backed study finds rare exposure to untrustworthy content on Facebook, Instagram

A collaboration between Meta Platforms Inc. and academic researchers published a study in *Science Advances* that examined the prevalence of content from repeatedly flagged “untrustworthy” sources on Facebook and Instagram during the 2020 United States election. The analysis covered roughly 231 million adult US Facebook accounts and 200 million Instagram accounts. On average, only about 1.1 % of the feed content seen by the median Facebook user and 0.1 % on Instagram came from these sources, but the exposure was highly concentrated: roughly 23 % of Facebook users (about 53 million accounts) received nearly 80 % of the untrustworthy content, and the top 2.5 % of users saw a majority of such material.

In a randomized three‑month experiment, posts from repeat offenders were removed from the feeds of half of nearly 16 000 consenting users. This cut exposure by about 70 %. Researchers tracked ten outcomes—including belief in false claims, ability to discern truth from falsehood, political polarization and trust in media—and found no measurable change in any metric, even among the most heavily exposed users.