How rumours and false news spread, and what works against them

Rumours are as old as human society, but the way they spread has changed. A claim can now reach millions of people within hours, crossing platforms, languages and borders before anyone has checked it. Research on rumours and false news has grown rapidly in response. This guide summarises what that research has found about why false information spreads and what helps against it.

Some definitions

Researchers Claire Wardle and Hossein Derakhshan proposed a widely used framework for information disorder in a report for the Council of Europe:

• Misinformation: false information shared without intent to harm. Example: someone forwards a fake health tip believing it is true.
• Disinformation: false information shared deliberately to deceive or cause harm.
• Malinformation: genuine information shared to cause harm, such as leaked private data.

A rumour is broader: an unverified claim that circulates. It may turn out to be true. Treating every rumour as false is itself a mistake, and an expensive one in a crisis.

Why rumours spread

The classic social-psychology account, from Gordon Allport and Leo Postman in the 1940s, is that rumours spread in proportion to the importance of a topic multiplied by its ambiguity. If something matters to people and official information is missing or distrusted, rumours fill the gap. This is why rumours flourish during disasters, epidemics and political crises.

Digital platforms add new drivers:

• Novelty and emotion. In the largest study of its kind, Soroush Vosoughi, Deb Roy and Sinan Aral analysed about 126,000 stories on Twitter between 2006 and 2017. False stories spread “farther, faster, deeper and more broadly” than true ones. False stories were more novel and provoked more surprise and disgust. People, not bots, were the main reason for the difference.
• Network structure. Claims travel fastest through tightly connected communities, where repetition from trusted contacts makes them feel familiar and therefore true. This is the illusory truth effect.
• Ranking systems. Algorithms that optimise for engagement can amplify outrage and novelty, whatever their accuracy.
• Low cost of production. Generative AI makes convincing fake text, images, audio and video cheap to produce at scale.

Why debunking sometimes fails

Correcting a false claim is harder than spreading it. Several findings explain why:

• The continued influence effect. People often keep relying on misinformation after they accept a correction, especially if nothing replaces the false explanation in their minds.
• Reach. Corrections rarely reach as many people as the original claim, and they reach them later.
• Trust. Corrections from sources the audience distrusts can be dismissed. In some studies they even harden beliefs, although the once-feared “backfire effect” turned out to be rarer than first thought.
• Repetition. Repeating a myth in order to debunk it can make it more familiar if it is done carelessly.

What works

The Debunking Handbook 2020, written by more than 20 researchers, summarises the evidence into practical advice.

Debunk with a “truth sandwich”:

1. Lead with the fact, stated clearly and simply.
2. Warn that a myth is coming, and mention it only once.
3. Explain why the myth is wrong and why it is misleading: who benefits, what technique it uses.
4. Finish by repeating the fact.

Prebunk where you can. Inoculation theory holds that exposing people to a weakened form of a manipulation technique helps them resist the real thing later. In a 2022 study, Jon Roozenbeek, Sander van der Linden and colleagues showed short videos explaining techniques such as emotional language, false dichotomies and scapegoating. The videos improved people’s ability to recognise manipulation, including when they were shown as YouTube ads to millions of real users. Prebunking works on techniques, so it does not require predicting each specific false claim.

Fill the information vacuum. In a crisis, the best antidote to rumours is fast, frequent and honest official communication, including about what is not yet known. Silence creates the ambiguity that rumours need.

Use trusted messengers. Local doctors, community leaders and people within the affected group are often more persuasive than national institutions.

Add friction, not just labels. Prompts that ask people to consider accuracy before sharing, and limits on mass forwarding, have reduced the spread of false content in some field experiments.

Monitoring rumours: an intelligence task

For governments, companies and newsrooms, rumour management starts with monitoring. That means tracking which claims are circulating, how fast they are growing, who is spreading them and in which communities. That is an intelligence task: it calls for clear questions, systematic collection, source evaluation and judgement about which rumours need a response. Not every false claim deserves a public correction. Responding to a fringe claim can spread it further. A common rule of thumb is to respond when a claim is reaching beyond its original community, or when it could cause real harm.

The road ahead

Two developments will define the next few years. Generative AI will make fabricated media more common and more convincing, which raises the value of provenance standards and verification skills. AI systems will also increasingly sit between people and information, answering questions directly instead of showing links. How those systems choose and cite their sources will shape what people believe. Research on rumours, which began with wartime gossip, is becoming research on how societies keep a shared base of facts.