“You won’t believe #7.” The line is familiar because the mechanism is familiar. A number implies order; a missing item implies a secret; the wording makes a small empty space in the mind and asks for a click to close it. Then the page opens. The seventh item is ordinary, the promise has dissolved, and the reader has paid attention for nothing.
That is not a failure of curiosity. Curiosity is useful. George Loewenstein’s 1994 information-gap account describes it plainly: we become curious when we can see the distance between what we know and what we want to know. A good headline can create that distance without hiding the road. “Why this checkout form loses mobile customers” gives the reader a reason to continue and a standard by which to judge the article. The article still has to do the work.
This is the line worth protecting: a headline is a contract. It can be sharp, specific, even provocative. It may promise a surprise, a conflict, or an inconvenient answer. But the page beneath it must repay the promise in substance, not merely gesture toward it after three screens of warmed-over setup. Clickbait begins where the gap is opened and left unpaid.
That definition matters because “clickbait” is too often used for any headline that tries to be interesting. It is not a moral category for people who click, nor a command to write blandly. A publisher watching only this week’s click-through rate has a rational reason to push harder. Negative emotion can pull attention fast. The question is whether the gain belongs to the article or merely to its packaging.
The Upworthy experiments make the usual advice look less tidy. In a PLOS One registered report published in 2023, researchers analyzed thousands of headline A/B tests. Longer headlines performed better in that dataset, and negative-emotion language was associated with more clicks. Positive-emotion language and readability were not significant. Language features alone predicted the winning headline only modestly—about 54.4% accuracy. Those are dataset findings, not an instruction to make every title longer or darker. Topic, audience, timing, placement, and the actual article still decide far more than a formula can.
This is where the industry’s “power words” lists become misleading. Emotion is not fake simply because it appears in a headline. Berger and Milkman’s 2012 research on online sharing found that high-arousal emotions can increase sharing; anger and anxiety can move people, but so can awe. The usable lesson is not that one adjective is a button marked more clicks. Emotional language has a modest, context-dependent effect. It should clarify the real stakes of a piece, not manufacture stakes the piece cannot support.
The same restraint applies to formulas. A swipe file can be a serious craft tool: keep headlines that make you stop, then study what they promised, what page they led to, and what kind of reader they addressed. It is a library of decisions, not a bag of spells. Headline analyzers can help catch a missing keyword or an accidentally opaque phrase, but their scores are not science and are not reliable predictors of real performance. A formula cannot make a promise true. Only the writer, and then the article, can do that.
Length is another area where a useful technical detail has hardened into folklore. In 2025–2026 SEO practice, Google’s search display is constrained by pixel width, not a fixed character count—roughly 600 pixels on desktop in common guidance. Fifty to sixty characters can be a reasonable approximation for average English text, but it is not a law. Wide capitals and long words run out of room sooner; a precise title that truncates slightly may still serve the reader better than a shorter, vaguer one. Test the rendered result, not a ruler made from characters.
That is also why the visible H1 and the SEO <title> do not need to be identical. The H1 is the page’s opening promise to a person who has arrived. The title tag must also work in a search result, where space and query language matter. They should agree in meaning, keyword, and level of honesty, but they can take different shapes. A useful keyword belongs in both when it describes the page naturally. Forcing it into every available slot produces the old SEO sound: a page speaking to a crawler while a person waits nearby.
Testing still matters, but it should test a real choice. Write two headlines that make distinct, defensible promises. Check whether each fits the content, whether it tells the right reader why the subject matters, and whether the page satisfies the expectation it creates. Then look at the result alongside engagement, return visits, conversions where relevant, and reader feedback—not a single score from a tool. The better headline is not automatically the one that gets the first click. It is the one that begins the right relationship.
There is a historical reason this is more than a clean-copy preference. In 2014, Facebook explicitly adjusted distribution against pages that were “clicky” but not “sticky”: headlines that won the click without holding attention. Upworthy, a major symbol of the curiosity-gap era, fell from about 87 million monthly unique visitors at the end of 2013 to roughly 20 million by November 2014. Platforms change, and that example does not prove a universal law. It does show what happens when the ledger eventually includes the attention spent after the click.
The calm judgment comes late because the temptation is real. Clickbait can win in the short term. It borrows the same curiosity and emotional force that an honest headline uses. The difference is not excitement versus restraint; it is whether the writer respects the reader’s ability to judge what was promised. A brand is built in that small interval between headline and first paragraph. The same discipline belongs in its visual language too, which is why logo design best practices are not separate from the question of trust.
Before publishing a headline, read it once as a contract and once as the person who has just clicked it. If the second reader would ask, “Where is the thing you sold me?”, the title is not finished. The work is not to lower the temperature. It is to make the promise worth opening.