Misinformation

Misinformation in politics is false or misleading information that circulates among voters and shapes what they believe about candidates, measures, and public problems. Research since 2016 shows that exposure to known false news sources is a small part of most Americans’ media diet and is concentrated among a small group of very engaged users, yet false beliefs can last for years after they are corrected. The strongest remedies tested so far reduce belief or sharing modestly, and each raises a fair question about who decides what counts as false.

In plain words: some of what people read about politics is wrong. Sometimes it is an honest mistake. Sometimes it is made up on purpose. Most people see less of it than headlines suggest, but a small group sees and shares a lot of it. Fixing a false belief is possible, but the fix often fades. And any plan to remove false content means someone has to be trusted to decide what is false. This page lays out what is known, what is argued, and what is still open.

The problem, defined precisely

Misinformation is a contested term, and much of the disagreement about its size comes from different definitions. The Council of Europe report by Claire Wardle and Hossein Derakhshan (2017) organized the field around “mis-, dis- and mal-information,” which it called “the three elements of information disorder”.⁠¹ This page uses the following working definitions.

Misinformation. False or misleading content shared without an established intent to deceive.

Disinformation. False or misleading content created or spread with intent to deceive.

Malinformation. Accurate information used to cause harm, for example private details released to damage someone.

Fake news source. In much of the research cited below, a website classified by researchers or fact-checkers as publishing fabricated stories that imitate news. Studies count exposure to such sources, not to individual false claims. That is a measurement choice, and it matters: a source-level measure misses false claims made by mainstream outlets, officials, or ordinary users, and can count accurate stories from a flagged site.

Misperception. A belief that is false or unsupported by the best available evidence. A misperception can exist without any specific misinformation causing it.

Correction (fact-check). Information given after exposure that states the claim is false and explains why.

Prebunking (inoculation). Teaching people, before exposure, to recognize the techniques used to mislead.

Three quantities are often confused: exposure (did a person see it), sharing (did they pass it on), and belief (do they hold it). Evidence about one does not establish the others.

Scale and evidence

Misinformation looks large or small depending on which of those quantities is measured, and the honest summary requires keeping them apart.

Diffusion. Soroush Vosoughi, Deb Roy, and Sinan Aral studied “~126,000 stories tweeted by ~3 million people more than 4.5 million times” on Twitter from 2006 to 2017, classified true or false using six independent fact-checking organizations. They found falsehood “diffused significantly farther, faster, deeper, and more broadly than the truth in all categories of information,” with stronger effects for false political news. They also found that automated accounts “accelerated the spread of true and false news at the same rate,” which implies people, not bots, drove the difference.⁠² The denominator here is verified, fact-checked stories, which are not a random sample of all news.

Concentration of exposure. Nir Grinberg and colleagues linked Twitter accounts to registered voters during the 2016 U.S. presidential election. “Only 1% of individuals accounted for 80% of fake news source exposures, and 0.1% accounted for nearly 80% of fake news sources shared.” Most political news exposure, for people across the spectrum, “still came from mainstream media outlets”.⁠³

Sharing. Andrew Guess, Jonathan Nagler, and Joshua Tucker linked a survey to respondents’ Facebook data in the 2016 campaign. Among respondents with linked data, 8.5% shared at least one article from a fake news domain, and “over 90%” shared none. Users over 65 “shared nearly seven times as many articles from fake news domains as the youngest age group,” an effect that held after controlling for partisanship and ideology. The same study found conservatives more likely to share such articles, noting that fake news domains in 2016 “were largely pro-Trump in orientation”.⁠⁴

Share of the media diet. Jennifer Allen and colleagues, measuring news consumption across television and online sources, estimated that “fake news comprises only 0.15% of Americans’ daily media diet.” News of any kind made up at most 14.2% of the daily media diet, and television accounted for roughly five times as much news consumption as online sources.⁠⁵

Public concern. Pew Research Center surveyed 6,127 U.S. adults from February 19 to March 4, 2019. Sixty-eight percent said made-up news and information greatly impacts Americans’ confidence in government institutions. Sixty-two percent of Republicans and Republican-leaning independents called made-up news a very big problem, compared with 40% of Democrats and Democratic leaners. Respondents most often said political leaders and their staff create a lot of made-up news (57%).⁠⁶

Trust in news. Gallup’s September 2025 telephone poll of 1,000 U.S. adults found 28% expressing a “great deal” or “fair amount” of trust in newspapers, television, and radio. When Gallup began asking in the 1970s, “between 68% and 72% of Americans expressed confidence in reporting.” In 2025 the figure was 8% among Republicans, 27% among independents, and 51% among Democrats.⁠⁷

Support for restriction. In a Pew survey of 5,115 U.S. adults from June 5 to 11, 2023, 55% supported the U.S. government taking steps to restrict false information online even if it limits freedom to publish or access information, up from 39% in 2018. Sixty-five percent said technology companies should take such steps, up from 56%. The partisan gap was large: 70% of Democrats and Democratic leaners supported government action, compared with 39% of Republicans and Republican leaners.⁠⁸

Taken together, these numbers do not support either “misinformation is everywhere” or “misinformation is trivial.” Exposure is small on average and concentrated. Concern is widespread. Trust in the institutions that might referee the question has fallen sharply, and unevenly across partisan groups.

Causes and competing explanations

Researchers offer several explanations for why misinformation spreads and persists, and they are not mutually exclusive.

Repetition and fluency. The illusory truth effect is the tendency to judge repeated statements as more true. Lisa Fazio and colleagues found that “illusory truth effects occurred even when participants knew better,” meaning prior knowledge did not reliably protect people.⁠⁹

Novelty and emotion. Vosoughi and colleagues found false news was more novel than true news and that false stories drew replies expressing fear, disgust, and surprise. They suggest people are more likely to share novel information.⁠²

Inattention rather than partisanship. Gordon Pennycook and colleagues found that headline truth had little effect on sharing intentions despite a large effect on accuracy judgments, and argue that “people often share misinformation because their attention is focused on factors other than accuracy.” They write that their results “challenge the popular claim that people value partisanship over accuracy”.⁠¹⁰ Other scholars give identity and motivated reasoning a larger role; this remains an active dispute.

Elite and media cues. Brendan Nyhan argues that corrections often fade because they are “overwhelmed by cues from elites and the media promoting more congenial but less accurate claims,” and that “misperceptions typically persist in public opinion for years” after debunking. He recommends reducing the link between group identities and false claims rather than only supplying more facts.⁠¹¹

Platform design. A common claim is that engagement-ranked feeds amplify false content. Randomized experiments with Meta during the 2020 U.S. election complicate this. Moving consenting users to chronological feeds substantially reduced their time on the platforms, and the amount of “political and untrustworthy content they saw increased,” yet the change “did not significantly alter levels of issue polarization” or political knowledge over three months.⁠¹² Removing reshared content reduced exposure to content from untrustworthy sources, but also produced “clear decreases in news knowledge” and did not significantly affect polarization or individual political attitudes.⁠¹³ These studies cover specific platforms, one election, and a limited period.

Demand and trust. A further explanation, consistent with the Gallup trend, is that low trust in news and government makes people more willing to accept alternative sources. The causal direction between low trust and misinformation is not settled. See Trust in elections and Local news collapse.

Who is harmed

Misinformation harms different people in different ways, and the evidence on each harm differs in strength.

Voters who act on false premises make choices they might not make with accurate information. The heavy users identified by Grinberg and Guess (older, highly engaged, more often on the political right in 2016) carry the largest exposure, which means harm is concentrated rather than spread evenly.⁠^3,4

People and groups named in false claims, including election workers and candidates, bear reputational and sometimes personal costs. See Intimidation for related concerns.

Public institutions lose a shared factual baseline. Pew’s 2019 finding that 68% think made-up news greatly affects confidence in government shows a perceived harm even where measured exposure is small.⁠⁶

Speakers can also be harmed by the response to misinformation. Content wrongly labeled false, or removed under pressure, restricts legitimate speech. In Murthy v. Missouri, states and individual users alleged that federal officials pressured platforms to suppress their speech about COVID-19 and elections.⁠¹⁹ Whether that harm occurred in that case was not decided on the merits (see below), but the risk is part of the problem, not outside it.

How much misinformation changes election outcomes is unknown. One experimental hint: across four countries, Ethan Porter and Thomas Wood found exposure to misinformation increased false beliefs by “less than 0.07 points” on a 5-point scale.⁠¹⁴ A single exposure in a survey is not the same as years of repeated exposure, so this does not settle the question.

What has been tried elsewhere

Several remedies have been tested, and each has shown both results and limits.

Fact-checking. In simultaneous experiments in Argentina, Nigeria, South Africa, and the United Kingdom covering 22 fact-checks, Porter and Wood found fact-checks reduced belief in misinformation “by at least 0.59 points” on a 5-point scale, with most effects “still apparent more than 2 wk later”.⁠¹⁴ Nyhan’s review finds corrections are “typically at least somewhat effective” but that effects “often do not last or accumulate”.⁠¹¹ The feared “backfire effect,” where corrections strengthen false beliefs, has not held up as a general pattern in that review.

Accuracy prompts. Pennycook and colleagues ran four survey experiments and a field experiment on Twitter and found that “subtly shifting attention to accuracy increases the quality of news” people share.⁠¹⁰

Prebunking. Jon Roozenbeek and colleagues produced five short videos explaining manipulation techniques such as false dichotomies and scapegoating. In six randomized studies (n = 6,464) and a YouTube field study (n = 22,632), the videos improved recognition of those techniques and the quality of sharing decisions, with effects “robust across the political spectrum”.⁠¹⁵

Crowd-sourced notes. Results on X’s Community Notes are mixed. A 2023 analysis by Yuwei Chuai and colleagues found “no evidence that the introduction of Community Notes significantly reduced engagement” with misleading tweets, and suggested notes may be too slow.⁠¹⁶ A 2024 preprint by an overlapping team, using 237,677 noted post cascades, found notes reduced the spread of misleading posts “by, on average, 62.0%” and increased the odds that authors deleted them by 103.4%, while again warning that notes “might be too slow to intervene in the early (and most viral) stage”.⁠¹⁷ The two studies measure different things (platform-wide rollout versus the effect of a displayed note), which helps explain the difference.

Platform policy reversal. In January 2025, Meta announced it would “end the current third party fact checking program in the United States” and “instead begin moving to a Community Notes program”.⁠¹⁸ That is a large natural experiment whose results have not yet been established in peer-reviewed work.

Government involvement. In Murthy v. Missouri (June 26, 2024), the Supreme Court held 6 to 3 that neither the state nor the individual plaintiffs had established Article III standing to seek an injunction against federal officials’ communications with platforms. Justice Barrett wrote for the Court; Justice Alito, joined by Justices Thomas and Gorsuch, dissented. The Court did not decide where government persuasion ends and unconstitutional coercion begins.⁠¹⁹

The strongest arguments across the political spectrum

Each major position on misinformation has a serious case, and each has a weak point its critics press.

The harm-reduction case. Proponents of stronger moderation argue that false claims about elections and public health cause real damage, that platforms already rank content and so are never neutral, and that labeling or reducing reach is a modest response. The fact-checking and prebunking evidence supports the claim that interventions can work.⁠^14,15 The weak point is the referee: when trust in news stands at 28%, and at 8% among Republicans, labels from those institutions may persuade the people who need them least.⁠⁷

The free-speech case. Critics of moderation, including many conservatives, libertarians, and civil libertarians on the left, argue that no authority is reliable enough to define political truth, that yesterday’s “misinformation” sometimes becomes tomorrow’s accepted view, and that government pressure on private platforms is censorship by proxy. The Murthy plaintiffs made that argument.⁠¹⁹ The weak point is that “more speech” remedies also have limits: Nyhan finds corrections fade.⁠¹¹

The market and crowd case. A related libertarian argument favors decentralized remedies such as crowd notes, user choice of feeds, and competition among sources, because they spread the power to judge truth instead of concentrating it. Meta’s 2025 shift adopts this logic.⁠¹⁸ The evidence that crowd notes act fast enough is mixed.⁠^16,17

The skeptical-empirical case. Some researchers argue the problem is overstated: exposure is 0.15% of the media diet, and most people see mostly mainstream news.⁠^5,3 Overstating the threat can itself erode trust and justify overbroad restrictions. The weak point is that averages hide concentration, and concentrated exposure among highly engaged people may matter more than its share suggests.

The institutionalist case. Others argue misinformation is a symptom of collapsed trust and weakened local reporting, so the remedy is rebuilding credible, accountable institutions rather than policing content. This view is hard to test quickly and offers little for the next election.

Unresolved questions

• How much does misinformation change votes or election outcomes, as distinct from beliefs measured in surveys?
• Exposure on encrypted messaging apps and video platforms is hard to measure; are the low averages from Twitter, Facebook, and web panels representative of those channels?
• Why do corrections fade, and can repeated or source-matched corrections produce lasting change?.⁠¹¹
• Who should decide what is false, and how should that decision be reviewed and appealed? The Murthy decision left the line between persuasion and coercion undefined⁠¹⁹
• What did ending third-party fact-checking on Meta’s U.S. platforms change, if anything, in measured exposure?⁠¹⁸
• Are low trust in news and belief in false claims a cause of each other, products of a third factor, or both?