What actually travels

You posted something true. It got a few shares. Your competitor posted something that turned out to be false. It went everywhere. You watch it happen and you wonder what you are doing wrong.

So you look for an explanation. Maybe the platform favors them. Maybe you need better timing. Maybe you need more followers to start with. Those all sound like things you could fix.

Here is what the largest study on this actually found. False news spreads faster, farther and deeper than true news. Not a little faster. A lot faster. And the reason is not bots. It is people. People share false stories more because those stories are more novel. And because they provoke stronger feelings.

The finding is not about lying. It is about what the share button responds to. Novelty and emotion, measured at scale.

That is one half of how something spreads. The other half is the platform. It decides who sees a post at all, and a new post starts cold, with no audience of its own. This post is about the human half. Every source below is linked, dated and quoted.

The study everyone cites, and what it actually measured

Three researchers at MIT analyzed every rumor cascade they could verify on Twitter. From 2006 to 2017. They published it in Science on 9 March 2018. The paper is "The spread of true and false news online" by Soroush Vosoughi, Deb Roy and Sinan Aral.

A rumor cascade is one claim and everyone who spread it. The cascade starts when someone posts the rumor. It grows as others share it. Each share branches off. That makes a tree shape. The researchers gathered about 126,000 of these trees. They covered roughly 3 million people.

Then they did the hard part. They sorted the rumors into true and false. They used six fact-checking organizations to do it. Snopes, Politifact, FactCheck.org, TruthOrFiction, Hoax-Slayer and Urban Legends. A rumor only got marked true or false if those checkers agreed on it. Everything else was thrown out.

That step is what makes the study unusual. Most research on misinformation guesses at what is false. Or it tracks one known hoax. This one verified every claim in the dataset. That took eleven years of Twitter data. It is the largest verified study of its kind.

What they measured is in the next sentence. How far a story went. How fast it got there. How many people it touched. And how many steps it took from the first post to the last one.

Falsehood diffused significantly farther, faster, deeper, and more broadly than the truth in all categories of information. — Vosoughi, Roy & Aral, Science, 2018

Farther means more people. Faster means less time to the same number. Deeper means more steps in the tree. Broader means more branches. False beat true on all four.

The numbers are not close

The gap is large enough to see without a chart. False news reached between 1,000 and 100,000 people. True news rarely went past 1,000. That is the range, not the average. The average is smaller but the same direction. False information reached 35% more people than true information did.

Speed tells the same story. A false story took about 10 hours to reach 1,500 people on Twitter. A true story took about 60 hours to reach the same number. Six times slower.

The researchers broke the rumors into categories. Politics, urban legends, business, terrorism, science, entertainment and natural disasters. False news spread farther and faster in every single category. Politics had the largest gap. But the gap was there in all of them.

The two headline gaps, in one place:

• Reach. False stories reached between 1,000 and 100,000 people.
  • True stories rarely went past 1,000.
• Speed, measured as time to 1,500 people. False stories took about 10 hours.
  • True stories took about 60 hours.

Source: Vosoughi, Roy & Aral, Science, 2018.

That is the result. Now comes the part everyone gets wrong. Why it happens.

Bots are not the reason

The first guess is always bots. Automated accounts spreading junk. The researchers tested for that. They built a bot-detection model. Then they removed the bots from the data. Then they ran the analysis again.

False news still spread farther, faster and deeper. Removing bots made no difference to the result. Humans were doing this. Not machines.

The paper is direct about it:

We found that false news was more novel than true news, which suggests that people were more likely to share novel information. — Vosoughi, Roy & Aral, Science, 2018

Novel means new to the person reading it. Something they have not seen before. The researchers measured novelty by comparing each rumor to what a user had seen in the past 60 days. False rumors were more different from that history than true ones were.

Then they looked at replies. What people said when they saw a true story versus a false one. They used sentiment analysis to score the emotions in those replies. Sentiment analysis is software that reads text and guesses what feeling it carries.

False stories inspired fear, disgust and surprise in the replies. True stories inspired a different pattern. The paper does not name what that pattern is. It only says false and true provoked different emotional responses. And the false ones were stronger.

So the mechanism is not bots pushing junk. It is people choosing to share things that feel new and make them feel something strong.

What this means for anything you post

The study is about what travels. Not about what is good. Not about what should travel. Just what does.

If you make something novel, it has a better chance. Novel here does not mean original in the world. It means new to the person seeing it. Something they have not encountered in their feed recently. Something that surprises them.

If you make something that provokes an emotional reaction, it has a better chance. The emotions the study names are fear, disgust and surprise. Those are not the only ones. They are the ones false news triggered. But the principle is broader. A strong feeling makes someone more likely to share.

That is the measured answer to why some things spread. It is not timing. It is not follower count. It is not the platform favoring someone. It is novelty and emotion. Measured across 126,000 cascades and 3 million people.

One warning goes with that. The study measured what spread. It did not measure what worked. A post going viral and a post achieving your goal are not the same thing. If your goal is reach for its own sake, this is the map. If your goal is something else, reach is one input to it. Not the whole answer.

What this study does not carry over to Instagram

The finding is about organic human behavior on Twitter. People choosing to share things with their own followers. For free. Because the thing felt novel or made them feel something. Two things stop it at the Twitter border.

The first is the platform. Twitter is built for retweets. One share sends a post to everyone who follows that person, at once. Instagram does not work that way. A like or a comment does not broadcast the post to anyone's network. It may signal to Instagram's ranking models that the post is worth showing more. That is not the same as a direct retweet.

The second is the mechanism. The paper measured unprompted sharing. Someone sees a thing and passes it on, because it felt new or made them feel something. It does not measure engagement that was asked for, paid for or arranged in any way. Whether early engagement of that kind causes a post to reach more people on Instagram is a separate question. No study we could find has measured it.

So the study gives you a map of what humans naturally share. It does not give you a map of Instagram.

What the study does not cover

Five things, named so no one assumes the researchers checked them.

Whether novelty and emotion work the same way on other platforms. The study is Twitter only. From 2006 to 2017. That is before Twitter changed its timeline from chronological to algorithmic. It is also before TikTok. Before Instagram Reels. Before every platform started using recommendation models to decide what you see. The mechanism the study found may still hold. But it was measured in a different environment.

Whether true stories could spread faster if they were more novel. The study compared true news to false news as they naturally occurred. It did not test what happens if you make true news more surprising or emotional. That would be a different experiment.

Whether novelty and emotion are enough on their own. The study found they correlate with spread. Correlation means two things move together. It does not prove one causes the other. There could be a third factor. The authors do not claim causation in the quotes above. They say false news was more novel. They suggest that may explain why people shared it more.

What the effect size is for emotion versus novelty. The paper reports that both matter. It does not publish a weight for one versus the other. So you cannot tell which one drives more of the result.

Whether the same pattern holds for things that are not news. The dataset is rumors. Claims about events, policies, people and facts. The study does not cover art, entertainment, tutorials or personal stories. Those may spread differently.

The one number that does not move

Go back to the top of this post. You posted something true. Your competitor posted something false. Theirs spread farther.

You now know why. It was more novel to the people who saw it. Or it made them feel something stronger. Or both. The platform did not favor them. The bots did not push it. People did. At scale. In a pattern that held across 126,000 verified cascades.

That does not tell you to lie. The study is not advice. It is a measurement. What it measures is what the share button responds to. You can make true things more novel. You can make true things provoke feeling. Those are options. They are not the only options. But they are the ones this study found.

The gap is not small. False news reached six times as many people as true news, in the same time. That is not a platform quirk. That is human behavior, recorded at the largest scale anyone has checked.

What we could not determine

Where novelty ends and emotion begins. The study found both correlate with spread. It does not separate their effects. So we cannot say how much of the result is novelty and how much is emotion. Both matter. We do not know the split.

Whether the effect holds outside Twitter. The dataset is Twitter only. The authors state that as a limit in the paper itself. We cannot extend the finding to Instagram, TikTok, Facebook or anywhere else without a study on those platforms.

Whether prompted engagement has the same effect as organic sharing. The study measured people sharing things because those things felt novel or emotional. Engagement that arrives because it was asked for is a different input, and the paper never looks at it. Whether it produces more organic reach later is unmeasured here. No study we found measures it anywhere.

What makes something feel novel to a given person. The researchers measured novelty by comparing each post to what a user saw in the past 60 days. That is one measure. It is not the only way something can feel new. A post could repeat an old claim in a new way. Or frame a familiar fact with a surprising angle. Those may feel novel even if the core claim is not new to the dataset.

Whether this pattern has changed since 2017. The data ends in 2017. That is before most platforms switched fully to algorithmic feeds. The study measured what spread when Twitter showed posts in time order. Algorithmic ranking may change the result. We do not know.

How we did this

Scope: this post reads one published study and reports what it measured. No data collection of our own. No independent verification of the claims in the study.

How we read it: the study was retrieved on 18 August 2026 as a PDF from a university-hosted source. The abstract and key findings were extracted and quoted verbatim. We did not access the full text behind the paywall at Science's own site. The PDF we read carries the full paper. The quotations above are from that document. The sample size, methods, findings and limitations are reported as the authors stated them.

Limits: we read one study. Other research on misinformation may find different results or use different methods. This study is about Twitter from 2006 to 2017. It may not generalize to other platforms or other time periods. The study is observational. It does not run an experiment where novelty or emotion is manipulated. So the causation is the authors' interpretation, not a proven mechanism.

Sources

1. Soroush Vosoughi, Deb Roy and Sinan Aral, The spread of true and false news online. Science 359(6380):1146-1151, 9 March 2018, doi:10.1126/science.aap9559. MIT Media Lab and MIT Sloan School of Management. Sample: approximately 126,000 rumor cascades spread by roughly 3 million people on Twitter, 2006 to 2017. Method: rumor verification by six independent fact-checking organizations; cascade analysis measuring reach, speed, depth and breadth; bot-detection model applied; sentiment analysis of replies. Key quantitative findings: false news reached 1,000 to 100,000 people while truth rarely exceeded 1,000; false stories took about 10 hours to reach 1,500 users versus about 60 hours for truth; false information reached 35% more people on average; false news was more novel and inspired fear, disgust and surprise while true news inspired different emotional patterns. Authors' stated limitation: "focused on Twitter data only and examined verified rumors, which may not represent all misinformation online." Full PDF retrieved 18 August 2026.

Changelog

• 2026-08-18: First version drafted.