Does labelling misinformation actually work?
Modestly, with conditions — and the research is more encouraging than the pessimistic framing usually suggests, while falling well short of solving the problem.
What the evidence supports:
Labels reduce belief and sharing, on average. Studies of warning labels on social media content generally find small but real reductions in perceived accuracy and in sharing intention.
Specific beats generic. Labels explaining why something is disputed, with a link to the correction, outperform vague warnings.
Prebunking works better than debunking. Inoculation theory — forewarning people about a manipulation technique before they encounter it — has reasonably good evidence. Brief interventions explaining common tactics have reduced susceptibility in controlled studies, and the effect generalises across topics rather than requiring a label per claim.
Accuracy prompts. Simply asking people to consider accuracy before sharing measurably reduces the spread of false content. Research by Gordon Pennycook and David Rand suggests much sharing of misinformation reflects inattention rather than sincere belief — people are distracted by engagement rather than deliberately misleading.
The genuine problems:
The implied truth effect. This is the most important finding. Labelling some false content causes unlabelled false content to be perceived as more accurate — because users infer that anything unlabelled has been checked. Since labelling everything is impossible, partial labelling creates a systematic vulnerability.
Backfire effects were once believed to be common — that correction entrenches belief. Later work has found them rare and hard to replicate, which is a rare case of the pessimistic finding being the one that failed.
Corrections reach fewer people than the original claim, consistently.
Continued influence. Even after accepting a correction, people continue to use the false information in reasoning.
Source distrust. Labels from a distrusted institution can be dismissed or read as confirmation.
Friction — adding a step before sharing — appears effective and is underused.