Question explored with the scientific record
Do fact-checks and prebunking actually change what people believe?
Fact-checks and prebunking can shift beliefs, but they do not erase the deeper problem: people learn more from information that confirms what they already believe.
The most direct evidence comes from a 2026 study on confirmation bias [2]. In two experiments, researchers gave people a learning task where some information was labeled "true" and some "false." Even with those labels, people still learned more from confirmatory evidence than from disconfirmatory evidence. The effect was large and statistically significant in both studies (p < 0.001 and p = 0.01) [2]. Explicit reliability cues helped a little, but they did not eliminate the asymmetry.
A 2026 narrative review on science communication [1] and a 2025 study on vaccination hesitancy [4] both point to the same conclusion. The hesitancy study found that exposure to misinformation was the strongest predictor of vaccine refusal (adjusted odds ratio 3.12, p < 0.001), but low trust in health authorities was nearly as strong (aOR 2.48, p < 0.001) [4]. That means fact-checks compete against a trust deficit that no single correction can fix.
A county-level study of COVID-19 vaccination rates found that political affiliation was the strongest predictor of being unvaccinated, while a Twitter misinformation variable showed no significant association [3]. That suggests that what drives refusal is not simply exposure to false claims, but identity and distrust.
| Factor | Effect on belief change |
|---|---|
| Confirmation bias | Strong, persistent [2] |
| Explicit "true/false" labels | Weak attenuation [2] |
| Misinformation exposure | Strong predictor of hesitancy [4] |
| Trust in authorities | Nearly as strong as misinformation [4] |
| Political identity | Strongest predictor of behavior [3] |
My call: fact-checks and prebunking can change what people believe, but their effect is limited by confirmation bias and eroded trust. The evidence shows these interventions are not a cure for misinformation. Confidence: moderate.
Sources used 4
-
Science communication about contentious biology topics: why it frequently fails and how to fix it
A narrative review of evidence-based strategies for communicating contested biological science to non-expert audiences, covering audience awareness, trust, prebunking, debunking, tone, narrative, and classroom implementation.
DOI: 10.1128/jmbe.00194-26 -
Confirmation Bias Exists in the Face of False Information
Explicit reliability cues attenuate learning but do not erase confirmation bias: learning remains asymmetrical for confirmatory versus disconfirmatory evidence with both true and false labels, best captured by a four-rate reinforcement-learning model.
DOI: 10.64898/2026.05.07.723487 -
A population level study on the determinants of COVID-19 vaccination rates at the U.S. county level
In a county-level analysis of the U.S., political affiliation (measured as Republican presidential vote percentage) had the strongest positive association with the percentage of the population unvaccinated against COVID-19, while median household income was the strongest negativ…
DOI: 10.1038/s41598-024-54441-x -
Vaccination Hesitancy Trends among Adults Post-Pandemic: A Multi-Center Study
In a two-center cross-sectional study of 235 adults, 49.8% exhibited some degree of post-pandemic vaccination hesitancy; older age, female sex, lower education, unemployment/retirement, prior COVID-19, misinformation exposure, and low trust in health authorities were associated …
DOI: 10.61336/ejcp/25-12-559