neur.ro

Clickbait (curiosity gap)

type: effect · origin: study · evidence: contested · themes: choices and costs

Claim assessed: Headlines that leave a curiosity gap attract more clicks than headlines that summarise the article.

Published · how we work

In short

Clickbait means headlines written to make you click, usually by leaving something unsaid: “What she found in the attic shocked her”, “Why you should stop doing this”. The idea behind it is the curiosity gap: when you know you are missing a piece of information, you feel the urge to find it out. Experiments comparing such headlines with headlines that summarise the article, however, point in every direction. The largest analysis so far, of almost 9,000 real tests, suggests that a headline that is too vague loses clicks just like one that is too concrete, and the differences are small.

What it says

The theory. Loewenstein (1994) proposed that curiosity arises when we notice a gap between what we know and what we would like to know. The gap has to be the right size: if we already know almost everything, we are no longer curious, and if we know almost nothing, the topic can put us off. Lab studies have confirmed this inverted-U shape: people are most curious about questions they feel half sure of. We haven’t read the theory or these studies; we describe them from Aubin Le Quéré and Matias (2025).

What direct experiments say. Aubin Le Quéré and Matias collected nine published experiments comparing “curiosity gap” headlines with headlines that summarise the article. Three found that the gap lowers clicks or the intention to read, two that it raises them, and four found no effect. Molina et al. (2021), for example, varied a single clickbait element in the headline of the same article and found no extra curiosity; they also observed that automated clickbait detectors agreed with each other only 47% of the time.

A website’s real tests. Upworthy, a US site that was very popular in 2013–2015, tested every article with several headlines shown at random to visitors. The archive published by Matias et al. (2021) contains 32,487 such experiments and more than 538 million impressions. Aubin Le Quéré and Matias (2025) used it in a Registered Report, with the analysis approved before seeing the data. They measured how concrete each headline is, that is, how easily you can picture the situation it describes, and analysed 8,977 tests:

For 8.7% of Upworthy’s tests, a more concrete headline would have done better; for 50.9%, a vaguer one. The authors conclude that what works is a gap of the right size, not the gap as such.

Estimated click-through rate for two headlines that differ in concreteness, according to Aubin Le Quéré and Matias. Among very vague headlines: the more concrete one 0.91%, the vaguer one 0.86%. Among very concrete headlines: the vaguer one 1.17%, the more concrete one 1.06%. Neither too vague nor too concrete Aubin Le Quéré and Matias (2025), Upworthy tests Among very vague headlines estimated click-through rate the more concrete headline 0.91% the vaguer headline 0.86% Among very concrete headlines estimated click-through rate the vaguer headline 1.17% the more concrete headline 1.06%
Model estimates from Aubin Le Quéré and Matias (2025) for headlines one standard deviation below and above the test mean, in the vaguest and the most concrete tests. The headline that moves toward the middle wins.

What else matters. Gligorić et al. (2023), also in a Registered Report on the Upworthy data, compared 24,333 pairs of headlines. Negative words, longer headlines and certain pronouns were associated with more clicks, but all the language features together picked the winning headline in only 54% of pairs, against 50% by pure guessing. Robertson et al. (2023), on the same data, found that each additional negative word raised the click-through rate by 2.3%. Kuiken et al. (2017), on 1,828 headlines rewritten by the editors of a Dutch online news kiosk, found that words that announce something (“what”, “why”) increased clicks, while questions decreased them; their data, however, don’t come from randomised tests.

And after the click? Munger et al. (2020), in two experiments, didn’t find that reading clickbait headlines changes trust in the media, polarisation or what people remember. Molyneux and Coddington (2019) suggest such headlines may lower perceived credibility. Qiu and Golman (2024), across more than 100,000 WeChat articles, found that headlines raising a question brought more clicks, but not more likes relative to clicks. We read these three studies only as abstracts.

Example

The same article about a new parking rule can be announced as “City introduces double fees in the centre from 1 November” or “The change that will cost drivers from November”. The second headline leaves a gap; according to the data above, whether it gets more clicks depends on how vague the headlines it competes with are. The example is an illustration, not a study cited here.

How to apply it

The steps below are a practical way to apply it that we propose, based on the studies cited.

  1. When you write a headline, aim for the middle. Say enough for the reader to understand what it’s about, but not everything. A completely vague headline and one that summarises the whole article both lose, according to the Upworthy data.
  2. Test, don’t guess. Even with all the language rules, the winning headline was picked only slightly better than chance. If you can, compare two versions on real readers.
  3. As a reader, notice the gap. When a headline promises information without giving it, ask whether you care about the topic or only about the gap. The quality of the article has nothing to do with how well the headline is built.

Limits and nuances

Sources

See also: Negativity bias, Zeigarnik effect