What is rage bait, and how do you stop taking it?

Summary: Rage bait is content built to make you angry because anger is the reaction that platforms reward most and that readers act on fastest. It works by naming an enemy, not by making a claim, which is why it spreads whether or not it's true. The way to stop taking it is to notice the target before you notice the topic, and to treat the urge to share as the signal that you've been handed bait.

Rage bait is built to make you feel something strongly enough to act within the next few seconds, and the action it wants is a share, a reply, or a comment, each of which the platform counts as a success and shows to more people. The content can be true, false, or too vague to be either, because the claim was never the point. The point is the enemy it names and the reaction that follows, and a reader who learns to see the enemy before the topic has already stopped taking most of it.

Where the word comes from

Oxford University Press, the publisher of the Oxford English Dictionary, named "rage bait" its Word of the Year on December 1, 2025, and defined it as "online content deliberately designed to elicit anger or outrage by being frustrating, provocative, or offensive." Its use had roughly tripled over the previous 12 months. The term is older than that: Oxford's lexicographers traced its first online appearance to a 2002 post on Usenet, an early internet discussion system, where it described a driver's reaction to being flashed by another car trying to pass. Casper Grathwohl, the president of Oxford Languages, said the word "shines a light on the content purposefully engineered to spark outrage and drive clicks."

The definition's useful word is "deliberately." Plenty of content makes people angry by accident, because the world contains things worth being angry about. Rage bait is the subset where the anger is the product, manufactured to a specification, and the specification is known.

What the specification is

Researchers have measured, in several large datasets, which features of a post predict how far it travels, and the answers are consistent. In 2017, the psychologist William Brady and colleagues at New York University analyzed 563,312 tweets about 3 polarizing political issues and found that each moral-emotional word in a message, words such as "shame," "evil," or "disgrace," increased its diffusion by about 20%, and that the increase happened mostly within ideological groups rather than across them.

In 2021, the psychologists Steve Rathje, Jay Van Bavel, and Sander van der Linden examined 2.73 million posts from American news outlets and members of Congress on Facebook and Twitter. Posts about the political out-group, the other side, were shared about twice as often as posts about the author's own side, and each additional out-group term increased the odds of a post being shared by 67%. Out-group language was the strongest predictor of sharing they measured, about 4.8 times as strong as negative emotional language in general and 6.7 times as strong as moral-emotional language. It was also a very strong predictor of "angry" reactions. The finding is the specification: Rage bait names the other side, and the naming does more work than any emotion word.

The platforms knew this and built it in. In October 2021, The Washington Post reported from Facebook's internal documents that, starting in 2017, the company's ranking formula treated emoji reactions, including the angry face, as 5 times as valuable as a like. A staffer had warned at the time that favoring controversial posts could open "the door to more spam/abuse/clickbait inadvertently." By 2019 the company's own data scientists had found that posts drawing angry reactions were "disproportionately likely to include misinformation, toxicity and low-quality news," and Facebook eventually set the weight of the angry reaction to zero. The algorithm had been paying 5 points for anger, and the content market had supplied it.

A 2025 audit found the same pattern on Twitter after its change of ownership. The computer scientist Smitha Milli and colleagues recruited 806 users in February 2023 and compared what the engagement-ranked feed showed them with a plain chronological feed. The ranked feed "amplifies emotionally charged, out-group hostile content," content that the users themselves said "makes them feel worse about their political out-group," and the users "do not prefer the political tweets selected by the algorithm." The feed gave people what they reacted to, which is a different thing from what they wanted.

Why it spreads whether or not it's true

Anger changes what a reader does before it changes what they believe. In 2024, a team of psychologists led by Killian McLoughlin, William Brady, and Molly Crockett published in the journal Science an analysis of 1,063,298 Facebook posts and 44,529 tweets, plus 2 experiments with about 1,500 participants. Misinformation drew more outrage than accurate news on both platforms, people were more likely to share a post the more anger reactions it had, and the more anger reactions a post had, the more willing people were to repost it without reading it first. Outraged readers were not worse at telling true from false when asked. They simply shared without asking.

A 2026 study in the journal Cognition and Emotion, summarized by PsyPost, a psychology news site, in April 2026, found the mechanism in the lab. Across 3 experiments, participants prompted to feel moral anger were significantly more willing to share headlines from low-credibility sources than participants prompted to feel disgust or nothing, and the authors concluded that anger "actively reduces a person's reliance on credibility when deciding to share information." That's rage bait's whole business model in a single sentence. The bait never has to convince you the source is good, as long as you're angry enough to stop checking.

The brain's part in this is the one we described in our article on negativity bias: Threats get priority processing, and a named enemy is the most legible threat there is.

The objection: Some things deserve anger

The strongest objection is that this framing lets the powerful off the hook. Injustice exists, the people responsible for it benefit when the public stays calm, and a reader trained to distrust their own anger is a reader who has been talked out of the one emotion that reliably produces political action. Calling something rage bait is also an easy way to dismiss a story you'd rather not engage with, and "don't share while angry" is advice that protects the status quo.

We'd concede most of that. Anger is a correct response to many true things, and the research above doesn't say otherwise; Brady's and Rathje's datasets were full of accurate posts that spread because they named a real wrong. The distinction rage bait depends on falls between a story that would still be worth your attention if it made you feel nothing and a story that has nothing left once the feeling drains out. A report that a company dumped waste into a river names a wrong, a victim, a document, and a date; It would be news in a flat voice. A post that a group you already dislike "wants to destroy" something names a target and a verb and stops. The first earns the anger. The second manufactures it, and the test for which one you're holding is whether anything remains when you take the target out.

How to stop taking it, in 6 checks

Each check looks at something on the screen, not at your own feelings, which is the point, since your feelings are what the bait is aimed at.

1. Find the target before the topic. Read the post once and name the group it wants you angry at. If you can name the group before you can name what they're supposed to have done, the group was the content. Rathje's out-group finding is why this check comes first.

2. Take the target out and read what's left. Delete the name of the enemy and the adjectives attached to it. If a checkable claim remains, with a who, a what, and a where, you're holding news that happens to be infuriating. If nothing remains, you're holding bait.

3. Look for the verb. Rage bait runs on verbs that describe intent rather than events: "Wants to," "is trying to," "would love to." Events have dates and witnesses. Intentions attributed to a group have neither, and can't be false.

4. Check whether the source gains from your reaction. An account that posts outrage about a different target every few hours is in the outrage business, and its product is your reply. Open the account's recent posts before engaging with any one of them. If every post names an enemy, the account is a factory.

5. Treat the urge to share as the alarm. McLoughlin's study found that anger makes people share without reading. Use that: The moment you reach for the share button, open the link instead and read to the end. Most bait doesn't survive its own article, which is often a few paragraphs that never deliver the headline's accusation.

6. Notice when your feed agrees with you. Brady's finding was that moral-emotional content spreads within a side, not across sides. If every outrage you see is aimed at the other side, you're seeing your side's bait supply rather than the world, a pattern our article on echo chambers and filter bubbles explains in detail.

What these checks can't tell you is whether the underlying claim is true, and they aren't meant to. A post can pass all 6 and still be wrong, and a true story can be dressed as bait. What they tell you is whether you're being asked to think or asked to react, and that's the decision the bait is trying to make for you.

Our weekday brief at informed.now has no comment section, no reactions, and no feed to rank them with. Each of its 6 summaries states what happened and cites the source behind it, so there's no target to find because none was planted. The checks above are for everywhere else.

The next time a post makes you angry enough to answer it, don't. Open the account that posted it and read its last 10 posts. If each one names a different enemy, you've found a factory, not a story, and the only move that costs it anything is to close the tab without leaving the reaction it was built to collect.