Recommended Free Tools
A 2024 Science study found that 2,107 registered U.S. voters—among a panel of 664,391 people active on Twitter during the August–November 2020 presidential-election period—accounted for 80% of the false or low-credibility news shared by the panel. Their posts reached 5.2% of registered voters on the platform. The group was disproportionately made up of women, older adults and registered Republicans, and its high volume appeared to come from manual, persistent retweeting rather than automation.
What the 2024 study actually measured
Sahar Baribi-Bartov, Briony Swire-Thompson and Nir Grinberg published the analysis in Science in 2024. The study followed registered U.S. voters who were active on Twitter during the 2020 presidential-election period and identified accounts that repeatedly shared material classified in the study as fake news or from low-credibility news sources.
As an Amazon Associate I earn from qualifying purchases.
The headline result concerns sharing: how often accounts redistributed the material. It is not a count of who created false stories, who merely saw them, or who believed them. The sample was limited to the United States, Twitter and that election period, so it should not be read as a rate for every social network or every kind of misinformation.
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Clear out junk files and repair common Windows errors3Scan for outdated or missing drivers - takes under a minute- The accounts were matched to registered voters rather than treated as anonymous, randomly selected internet users.
- The study examined a defined set of false or low-credibility news sources; changing that definition could change the result.
- The analysis describes an extreme concentration of activity inside this panel, not proof that a fixed percentage of all social-media users behaves the same way.
Who the “supersharers” were
The strongest demographic pattern was overrepresentation of women, older adults and people registered as Republicans. Secondary descriptions of the study identify the most overrepresented subgroup as middle-aged white Republican women living in Arizona, Florida and Texas.
#1 Best Overall
That is a pattern in this particular U.S. election sample, not a statement about every woman, older person, Republican or resident of those states. Demographic overrepresentation also does not establish that identity caused the behavior. The study did not show that these people formed one centrally directed organization.
They were not shown to be a bot network
The study reported that the volume appeared to be generated through manual and persistent retweeting. In other words, the accounts repeatedly chose to repost material at a very high rate; the analysis did not present the activity as automated posting.
Rank #2
“Manual” does not mean every account acted independently, and it does not establish what any individual believed. The evidence supports a description of unusually active human accounts, not a conclusion that they were bots, foreign agents or members of a single conspiracy.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →How the concentration compares with other studies
Earlier work found similarly uneven distributions, but the studies used different samples, platforms, periods and definitions. Their percentages therefore cannot be combined into one universal misinformation rate.
Rank #3
| Study | Setting | What was measured | Reported concentration |
|---|---|---|---|
| Science, 2019 | U.S. 2016 election | Twitter news consumption and sharing | Fake news was nearly 6% of Twitter news consumption; 1% of users accounted for 80% of exposures, while 0.1% accounted for nearly 80% of sharing. |
| Scientific Reports, 2022 | Twitter COVID-19 dataset | Creation and consumption of fake content | About 14% of users were classified as creators and 86% as consumers; the creator minority originated 82% of fake content in that dataset. |
The comparison suggests a recurring pattern: a small, highly active minority can shape a much larger audience’s information environment. It does not show that the same people, percentages or mechanisms apply across elections, health crises and platforms.
Why do people share false stories?
Habit can override evaluation
A 2023 PNAS study of 2,476 participants found that 30–40% of false news shared in its experiments was attributable to the 15% of participants who were the most habitual news sharers. Those habitual sharers often distributed both accurate and inaccurate material, which points to an automatic response to platform cues rather than a simple explanation based on ideological conviction.
Inattention is different from deliberate deception
A 2021 MIT summary of research on accuracy prompts separated false-headline sharing into three categories in its experimental setting: about 50% was linked to inattention, 33% to mistaken judgments that the headlines were accurate, and 16% to knowingly sharing material recognized as false.
Those percentages describe that experiment, not a census of all online misinformation. They do, however, show why treating every false share as proof of belief or malicious intent is misleading.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Can an accuracy reminder help?
In the accuracy-prompt research, asking people to think about whether a headline was accurate made them more discerning in what they shared, regardless of ideology. The finding supports a modest intervention: a brief prompt can interrupt automatic sharing and improve selection in an experimental setting.
It is not a guarantee that prompts will stop misinformation. Their effect depends on implementation, user attention, the material being evaluated and the surrounding platform incentives. Other responses may include reducing frictionless mass retweeting, improving source context and investing in media-literacy education, but no single measure eliminates false content.
What the finding does—and does not—tell us
- It does tell us: sharing was extraordinarily concentrated among a tiny group in the 2020 U.S. Twitter panel.
- It does not tell us: that 2,107 people produced all misinformation online, or that the result applies unchanged to Facebook, TikTok, YouTube or private messaging.
- It does tell us: the identified accounts were associated with particular demographic groups more often than the rest of the panel.
- It does not tell us: that demographic identity makes someone deceptive, that every member of a group shared false news, or that demographics alone explain the behavior.
- It does tell us: researchers observed manual, persistent retweeting.
- It does not tell us: whether every supersharer believed the material, coordinated with others or intended to mislead.
The practical lesson
When a false story appears ubiquitous, its apparent popularity may be produced by a very small number of highly active accounts rather than broad agreement. For readers, pausing to check the source and the claim before reposting addresses the inattention and habit mechanisms identified in the studies. For platforms, the evidence favors interventions that slow automatic amplification and make accuracy easier to consider.
Free tools Windows power users keep installed
One-click scans. No signup required.
The most defensible conclusion is narrow but important: during the 2020 U.S. election on Twitter, a tiny, demographically distinctive group of human users generated most of the false-news sharing observed in a large voter panel. That is a warning about concentrated amplification—not a license to stereotype an entire demographic or to treat one election study as a map of misinformation everywhere.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
