Fake News Statistics 2026: Spread, Trust and AI Content Farms

A landmark study from the MIT Media Lab, published in Science, set the baseline for how false information moves online. False news travelled significantly farther, faster, deeper, and more broadly than verified stories across approximately 126,000 cases on Twitter, shared more than 4.5 million times between 2006 and 2017. Seven years later, the World Economic Forum surveyed over 900 experts and ranked misinformation and disinformation as the number one short-term global risk for the second consecutive year.

Two unique angles cut through the noise here. Detection is not the bottleneck. Older adults outperform younger users at spotting fake news in lab tests, yet share it seven times more. Second, the AI-deepfake election apocalypse failed to arrive: less than 1% of fact-checked 2024 election misinformation was AI-generated, while NewsGuard’s industrialized AI content farm tracker grew from 2,089 to 3,006 sites in five months.

The data below covers spread mechanics, public trust, AI content farms, deepfake detection, demographics, and the economic cost of fake news.

Key Takeaways

• Per the MIT Media Lab analysis, the top 1% of false news cascades on Twitter reached between 1,000 and 100,000 people, while the truth rarely diffused to more than 1,000 people, drawn from approximately 126,000 stories.
• According to the Reuters Institute for the Study of Journalism, global trust in news sits at 40%, with Finland highest at 67% and Hungary and Greece tied at the bottom on 22%.
• About nine-in-ten US adults (90%) say they at least sometimes encounter inaccurate news, and 42% say it happens extremely often or often, according to Pew Research Center.
• NewsGuard had tracked at least 3,006 AI-generated content farm sites across 16 languages as of March 2026, up from at least 2,089 in October 2025, per NewsGuard’s AI tracker.
• Disinformation costs the global economy an estimated $78 billion a year, including $39 billion in stock-market losses, per a University of Baltimore and CHEQ study.
• According to research by Guess, Nagler and Tucker published in Science Advances, US Facebook users over 65 shared almost seven times more fake news than users under 30 during the 2016 election.

Editor’s Choice

• The MIT study analyzed approximately 126,000 stories tweeted by approximately 3 million people more than 4.5 million times.
• Reuters Institute fielding: 58% of respondents worldwide say they worry about distinguishing real from fake news online.
• Trust in US national news organizations has fallen by 20 points since 2016, with 56% still expressing some trust as of September 2025, per Pew Research Center.
• Deepfake files in circulation grew from about 500,000 in 2023 to about 8 million in 2025, an exponential expansion.
• Bot-driven amplification accounted for an estimated 23% of political discourse on X during election seasons, up from 15% in 2020, according to the Oxford Internet Institute.
• Per WHO, vaccines have saved more than 150 million lives over 50 years, yet 14.5 million infants missed essential immunization in 2024, partly attributed to vaccine misinformation.

Recent Developments

• March 2026: NewsGuard’s AI content-farm tracker identified at least 3,006 AI Content Farm websites across 16 languages.
• January 2025: Meta announced the end of its third-party fact-checking program in the United States, a switch to a Community Notes model on Facebook, Instagram, and Threads, and a revised estimate that approximately 3% of Facebook’s monthly active users are fake accounts.
• January 2025: The World Economic Forum’s Global Risks Report 2025, surveying over 900 experts and leaders, again ranked misinformation and disinformation as the #1 short-term global risk.
• December 2024: Romania’s Constitutional Court annulled the first-round results of its presidential election after declassified intelligence linked the leading candidate’s social-media campaign to a foreign-state operation.
• October 2025: Pew Research Center reported that 52% of Americans find it difficult to determine what is true and what is not when getting election news.

For a broader context on how attention defaults to novelty over verification, see SQ Magazine’s Google search statistics.

How Fast False News Spreads on Social Media

• The MIT Media Lab analysis: approximately 126,000 distinct stories tweeted by approximately 3 million people more than 4.5 million times between 2006 and 2017.
• Across approximately 126,000 stories shared more than 4.5 million times, fact-checker classification agreement ranged from 95 to 98%, drawn from six independent organizations.
• The top 1% of false news cascades reached between 1,000 and 100,000 people; verified true stories rarely reached more than 1,000, across approximately 126,000 stories analyzed.
• Falsehood diffused significantly farther, faster, deeper, and more broadly than the truth across every category studied within the approximately 126,000-story dataset shared more than 4.5 million times.
• The diffusion gap was most pronounced for false political news, exceeding the gap for terrorism, science, finance, natural disasters, and urban legends among the approximately 126,000 stories shared more than 4.5 million times.
• Removing all identified bots from the approximately 126,000-story dataset (shared more than 4.5 million times) did not eliminate the spread differential, indicating humans drive most of the gap.
• The researchers attributed faster sharing to novelty: across approximately 126,000 stories shared more than 4.5 million times, false stories were measurably more novel than true stories, and humans preferentially share novel information.

Source: Vosoughi, Roy, and Aral, Science (2018).

By the numbers: According to the MIT Media Lab analysis published in Science, falsehood diffused significantly farther, faster, deeper, and more broadly than the truth across all categories of information, with the top 1% of false cascades reaching between 1,000 and 100,000 people. The pattern persisted after bots were removed, pointing to human behavior as the principal amplifier.

For deeper context on how short attention windows and novelty preference reshape spread patterns, see SQ Magazine’s attention span statistics.

Don’t chase tech news. We track it for you.

One weekly briefing with the launches, AI developments, and breaches that matter. No filler.

Email address

Subscribe Now

Public Trust in News and Media Globally

• Per the Reuters Institute Digital News Report 2025, the survey provided record-high coverage of 48 markets across six continents.
• Overall trust in news sits at 40%, holding stable for the third consecutive year.
• Finland has one of the highest trust scores at 67%; Hungary and Greece share the lowest at 22%.
• 58% of respondents worldwide say they are worried about distinguishing real from fake news online.
• Concern about misinformation is highest in Africa and the United States, both at 73%, and lowest in Europe at 54%.
• 47% of respondents named online influencers and personalities as a top source of false or misleading information, tied with national politicians at 47%.
• News media and journalists themselves were named as a major misinformation threat by 32% of respondents.

Source: Reuters Institute Digital News Report 2025; Pew Research Center.

Fake News Exposure in the United States

• About nine in ten US adults (90%) say they at least sometimes come across news they believe is inaccurate, per Pew Research Center’s October 2025 short read.
• 42% say they encounter inaccurate news extremely often or often.
• 73% of US adults say they have seen inaccurate election news at least somewhat often, and 37% say they have seen it extremely or very often.
• 52% of Americans say it is generally difficult to determine what is true and what is not when getting news about elections.
• Trust in national news organizations stood at 56% in September 2025, down 20 points since Pew first asked the question in 2016.
• Among those who follow the news most closely, 46% say it frequently makes them angry.

Misinformation as a Global Systemic Risk

• The World Economic Forum’s Global Risks Report 2025 surveyed over 900 global risks experts, policymakers, and industry leaders between September and October 2024.
• Misinformation and disinformation ranked as the #1 risk over a two-year horizon for the second consecutive year.
• Extreme weather events ranked #2 over the two-year horizon and #1 over a 10-year horizon.
• Among the over 900 respondents, adverse outcomes of AI were named as a leading factor accelerating the spread of misleading video, images, voice, and text.
• The report notes that the volume of false or misleading content continues to rise, while the difficulty of distinguishing truth grows for citizens, companies, and governments alike.

Across our cybersecurity coverage, the consistent pattern is that breach costs and threat volumes climb faster than the budgets and headcount aimed at containing them. The misinformation curve looks similar: detection tooling improves yearly, yet the 2025 WEF survey still placed disinformation atop the short-term risk list for the second year running.

For broader systemic threat data, see SQ Magazine’s cybersecurity threat data.

AI-Generated Content Farms and Synthetic News

• Per NewsGuard’s AI Tracking Center: as of March 2026, NewsGuard’s team has identified 3,006 AI Content Farm news and information websites, up from at least 2,089 in October 2025.
• These at least 3,006 sites span 16 languages: Arabic, Chinese, Czech, Dutch, English, French, German, Indonesian, Italian, Korean, Portuguese, Russian, Spanish, Tagalog, Thai, and Turkish.
• Across at least 3,006 tracked sites, NewsGuard classifies a site as an Unreliable AI-Generated News Site (UAIN) when there is clear evidence that a substantial portion of content is produced by AI, with strong evidence of minimal human editorial oversight and no transparent disclosure to readers.
• Less than 1% of all fact-checked misinformation during the 2024 election cycles was AI-generated content, according to research summarized by Brookings.
• The growth of at least 917 sites in five months indicates the dominant AI misinformation risk has shifted from individual viral deepfakes toward industrialized content production.

For related synthetic-content risks, see SQ Magazine’s LLM safety statistics.

Deepfake Volume and Detection Statistics

• Deepfake files in circulation grew from about 500,000 in 2023 to about 8 million in 2025.
• Average human deepfake detection accuracy across modalities sits at 55.54%, barely above chance, according to a 2024 meta-analysis of 56 studies.
• Human detection rates for high-quality deepfake video are reported at 24.5%.
• Only 0.1% of participants correctly identified all fake and real media shown in a 2025 iProov study.
• Financial losses from deepfake-enabled fraud exceeded $200 million during the first quarter of 2025.

Source: iProov; meta-analysis of 56 deepfake studies.

The Economic Cost of Fake News

• Disinformation costs the global economy an estimated $78 billion a year, according to a University of Baltimore and CHEQ study published in November 2019.
• The study attributes $39 billion to stock-market losses tied to false information events.
• An additional $17 billion is attributed to poor financial decisions resulting from disinformation.
• Health misinformation accounts for an estimated $9 billion in annual business losses; reputation management adds another $9 billion.
• Platform safety efforts cost businesses about $3 billion a year, while fake political advertisements account for an estimated $400 million.
• The figures were derived from CHEQ’s proprietary ad-fraud and bot data, public stock-market data, and news reports analyzed by Professor Roberto Cavazos.

Key finding: Per the University of Baltimore and CHEQ 2019 economic study, fake news costs the global economy an estimated $78 billion a year, with $39 billion attributable to stock-market losses and $17 billion to poor financial decisions. The figure is now widely cited by the World Economic Forum, regulators, and corporate disinformation working groups when modelling disinformation risk exposure.

Demographics: Who Shares Fake News

• Per Guess, Nagler, and Tucker (2019) in Science Advances: during the 2016 US presidential election, only 3% of Facebook users under age 30 (those aged 18 to 29) shared links from fake news sites, compared with 11% of those over age 65.
• Adults over 65 shared almost seven times more fake news than adults under 30 in the same dataset.
• Comparing adults over 65 with adults under 30, holding education, ideology, and partisanship constant, the age effect remained the most robust predictor of sharing in the study.
• A meta-analysis of 31 studies concluded that older adults (those over 65) are better than young adults (under 30) at spotting fake news in laboratory settings.
• Comparing users over 65 with users under 30, the sharing gap therefore appears tied to congeniality bias and partisan motivated reasoning, not to detection ability.

The framing matters. Most fake-news roundups treat sharing volume as a literacy problem, with media-literacy programs as the headline fix. The research base now points the other way: detection is roughly fine, especially for older readers; the lever that actually moves is whether a story flatters one’s political tribe. Interventions that target congeniality bias, friction at the share button, and partisan-cue prompts are likelier to change behaviour than another quiz on spotting a fake URL.

Source: Guess, Nagler, Tucker, Science Advances (2019).

For a contrasting demographic on younger-user behaviour, see SQ Magazine’s Gen Z social media data.

Health Misinformation and Vaccine Hesitancy

• According to WHO: vaccines have saved more than 150 million lives in the past 50 years.
• In 2024, 14.5 million infants worldwide missed even a single dose of vaccine in the essential immunization programme, per WHO; vaccines have saved more than 150 million lives over 50 years.
• WHO attributes part of the coverage drop, against a backdrop of more than 150 million lives saved by vaccines, to misinformation, including the disproven claim that the MMR vaccine causes autism.
• Measles cases and deaths are rising in wealthy countries, including the United States, Canada, the United Kingdom, and other European countries, undermining the more than 150 million lives saved by vaccines over 50 years.
• WHO describes the COVID-era information environment as an infodemic, defined as an abundance of information, some accurate and some not, that occurs during an epidemic and erodes trust in health authorities responsible for the more than 150 million lives saved by vaccines.

Source: World Health Organization

Bot Amplification and Computational Propaganda

• Oxford Internet Institute history: the Programme on Democracy and Technology has been investigating computational propaganda since 2012.
• An OII study estimated that 23% of political discourse on X during election seasons involved bot-driven amplification, up from 15% in 2020.
• OII research documents the proliferation of state-linked cyber troops in countries around the world, the reach of state-sponsored media outlets, and the impacts of foreign influence operations.
• The institute frames computational propaganda as a tool deployed by states to shape public opinion, sow distrust, and create confusion both at home and abroad.

Source: Oxford Internet Institute, Programme on Democracy and Technology.

Election Misinformation and Voter Concern

• 87% of respondents in cross-country polling expressed concern about disinformation’s impact on upcoming elections in their country, with 47% describing themselves as “very concerned,” per data summarized by Brookings.
• 64% of election officials reported in 2022 that the spread of false information had made their jobs more dangerous.
• Less than 1% of all fact-checked misinformation during the 2024 election cycles was AI-generated content.
• In December 2024, Romania’s Constitutional Court annulled the first-round results of its presidential election after declassified intelligence linked the leading candidate’s social-media campaign to a foreign-state operation.
• Polling cited by Brookings indicates that false claims affected how voters viewed candidates and shaped opinions on the economy, immigration, and crime during the 2024 US cycle.

Platform Moderation and Content Removal

• Meta removed millions of pieces of content every day in December 2024, with fake accounts approximately 3% of Facebook MAU, per the company’s January 2025 enforcement update.
• Meta estimated that one to two out of every ten of those removals may have been mistakes, with fake accounts now approximately 3% of MAU, meaning the content likely did not actually violate platform policies.
• In January 2025, Meta announced it would end its third-party fact-checking program in the United States and shift to a Community Notes model on Facebook, Instagram, and Threads, alongside an estimate that fake accounts make up approximately 3% of MAU.
• Meta now estimates fake accounts at approximately 3% of Facebook monthly active users, lower than its previous 5% benchmark.
• Meta has committed to expanding transparency reporting on enforcement mistakes (alongside its approximately 3% fake-account estimate) so users can track changes over time.

Source: Meta Newsroom.

For a platform-by-platform view of misinformation moderation, see SQ Magazine’s social-media misinformation data.

Frequently Asked Questions (FAQs)

Conclusion

Fake news statistics keep landing in the same place. False stories travel further and faster than verified ones across approximately 126,000 cases shared more than 4.5 million times, with the top 1% of false cascades reaching up to 100,000 people while true ones rarely clear 1,000. Surveying over 900 experts, the World Economic Forum still ranks misinformation as the number one short-term global risk for the second consecutive year. Across 48 markets, 58% of news consumers say they worry about distinguishing real from false news online.

The shape of the problem is changing in this year’s data: less than 1% of 2024 fact-checks involved AI-generated content, undermining the predicted AI-deepfake election apocalypse. NewsGuard’s index of industrialized AI content farms reached at least 3,006 sites in 16 languages by March of this year. Researchers, regulators, advertisers, and election officials all benefit from sharper figures here; readers benefit from knowing which interventions, including friction at the share button and platform-level transparency, are most likely to move the needle.