Study of 4.5M+ tweets from 2006 to 2017 confirms inaccurate news spreads faster and further than true stories, and humans, not bots, are primarily to blame
An analysis of 4.5 million tweets shows falsehoods are 70 percent more likely to get shared — There's been a lot of talk …
Context & Ripple Effects
For most of 2018 the misinformation story was a bot story: Pew found the 500 most-active suspected bot accounts drove 22% of shared news-site links, and a later analysis put bots at 34% of shares of low-credibility articles despite being just 6% of accounts (14M-tweet analysis). This new Science study of 4.5M tweets cuts against that framing: falsehoods are 70% more likely to be retweeted, and the acceleration is attributed to human choices, not automation.
That reframing matters because it lands between two poles in the corpus — the bot-attribution studies and the finding that fake-news exposure was extraordinarily concentrated, with 0.1% of users accounting for nearly 80% of shares in 2016. If humans are the engine, the fix has to target sharing behavior rather than account authenticity.
First-order effects
- Twitter's enforcement posture, built around purging suspicious accounts, now addresses the smaller share of the problem — the study implies removing bots would not have stopped falsehoods from outrunning true stories.
- Researchers and platforms get a causal benchmark: novelty-driven human sharing, not coordinated automation, explains why false news travels faster and further.
Second-order effects
- Intervention design shifts toward friction at the moment of sharing — prompts, accuracy nudges, and feed-ranking changes aimed at people — because bot-detection tooling no longer covers the primary vector.
- Fact-checking organizations face a volume problem the corpus already quantifies: with exposure concentrated among a tiny sliver of heavy sharers, corrections must reach those users specifically or miss most of the audience for any given falsehood.
Third-order effects
- If human psychology is the dominant amplifier, platform accountability debates move from 'did you remove bad actors' to 'does your product design reward impulsive sharing' — a shift that invites regulatory scrutiny of engagement mechanics themselves.
- The pattern points toward a durable arms race in which authenticity infrastructure (provenance, verification) matters less than slowing the share button, since supply-side takedowns leave the demand side untouched.
The trend: Misinformation research is migrating from blaming automated accounts to explaining human sharing behavior, forcing platforms to redesign the act of sharing itself.