How to Write Headlines That Get Clicks (Without Being Clickbait)

Key Takeaways

• A clickable headline is specific, plain and front-loaded, and it promises only what the page delivers.
• People scan before they read. In Nielsen Norman Group's early web study, 79% of test users always scanned a new page and only 16% read word by word, so the first words of a headline carry the most weight.
• Facebook's 2017 definition of clickbait has two parts: headlines that withhold key facts and headlines that exaggerate. Avoid both and you are most of the way to an honest, clickable headline.
• Testing helps, but read the claims with care. A small Dutch newsroom reported 11% more clicks from reordering one headline, and the figure comes with no sample size.
• Write several versions, choose with a stated rule, then check that the first paragraph pays off the promise.

Why the headline does so much work

Most readers meet your headline before they meet anything else you wrote. It shows up in a feed, a search result, an email subject line or a link preview, often with no context around it. If it does not make sense on its own, nobody gets to the good part.

Nielsen Norman Group (NN/g) has been saying this for a long time. In a 1997 article, the group reported that 79% of its test users always scanned any new page and only 16% read word by word. That was a study of a different web, so treat the exact percentages as dated. The direction still matches what people do in a feed today: glance, judge, move on.

The same article tested five versions of one website with the same information but different wording. Against a promotional-style control, concise text scored 58% better on usability, a scannable layout 47% better, and neutral, objective language 27% better. All three together scored 124% better. The study did not publish participant numbers, so this is a pointer rather than a law, but it supports a plain rule for headlines: say the thing, briefly, without inflating it.

Four traits of a headline that earns the click

NN/g's 2017 guidance on microcontent, written by Hoa Loranger and Jakob Nielsen, covers headlines, page titles and subject lines. It boils down to four habits.

1. Be specific

Generic headlines get ignored. Useful ones carry a fact or a concrete detail that sparks interest. Compare "Tips to improve your marketing" with "Three subject-line changes to try this month" (both invented for illustration). The second names a thing the reader can picture. Only use a number if you really have one, and make sure the page backs it up.

2. Make it stand alone

Headlines often appear out of context in news feeds, social streams and blog listings. NN/g's advice is that a headline should make sense without the article next to it. Quick test: show it to someone who has never seen your site. If they cannot say what the page is about, rewrite it.

3. Put the key phrase first

Scanners read the start of a line and often stop there. NN/g's example reorders a health headline so that the topic leads and the surprise follows. Apply the same idea to yours: open with the words your reader is already thinking about, then add the twist.

4. Use plain language

The same guidance warns against puns and cute wording. A pun can make you smile in the draft and cost you clicks, because the reader has to decode it before deciding. Save the wit for the body copy.

What clickbait actually is

Clickbait gets used as a vague insult, so it helps to have a working definition. In a May 2017 update, Facebook described clickbait as headlines that either withhold or exaggerate. A withholding headline hides or misleads about the key fact so you have to click to learn it. An exaggerating headline uses sensational wording to make a story sound bigger than it is. Facebook said it found these by having reviewers label large sets of headlines and then looking for phrases that turn up often in clickbait and rarely elsewhere, much like spam filtering.

NN/g adds the business reason to avoid it: teasers that fail to deliver erode trust, and readers who are disappointed repeatedly stop coming back.

A "curiosity gap" is fine as long as the page closes it quickly. If the headline raises a question, answer it in the first few lines, not at the bottom after three scrolls.

Drafting prompts, not magic formulas

Formulas are useful for getting unstuck. They are not proof that a headline will work. Use these to produce a pile of drafts, then cut with the four traits above.

1. How to: How to [get the outcome] without [the usual pain].
2. List: [Number] [specific things] that [result].
3. Question: Are you making these [number] [topic] mistakes?
4. Mistake or warning: Stop doing [common habit] if you want [outcome].

For any of these, the discipline is the same. Fill the brackets with real details from your product or your data. If you cannot fill them, the content may not be ready, and the fix is upstream of the headline. Settling what your product promises and to whom, which is what positioning work is for, makes the brackets easy to fill.

Testing headlines without fooling yourself

Publishers test headlines routinely. Upworthy, the viral-content site, ran an A/B test on every article it published between January 2013 and April 2015, according to a 2021 paper in Scientific Data that released the test records. Each test averaged about five headline-and-image "packages," and each visitor session was randomly assigned one. The paper also says editors watched a dashboard, decided when to stop a test, and sometimes picked a package that was not the top performer. It reports that Upworthy did not run power analysis. In a 2024 update, the archive's maintainers flagged about 22% of the tests as affected by randomization problems and advised against using one period for confirmatory research.

A separate interview study of US newsrooms, by Nick Hagar and Nicholas Diakopoulos (Media and Communication, 2019), describes teams testing as many as a dozen headlines per article, with an algorithm picking the best one on a metric such as click-through rate.

Case study: 1Almere. 1Almere, a local news outlet in the Dutch city of Almere, tested two versions of one homepage headline with half of its visitors seeing each, while the image, text and layout stayed the same. One version led with the action, a school starting an investigation, instead of the person at the centre of the story. According to the write-up published by the CMS vendor Prepr, the action-first version drew 11% more clicks from homepage readers. The tests ran for a few days, and the article gives no sample size or significance level. It comes from a company that sells testing tools, so read 11% as an encouraging single example, not a benchmark.

The lesson from the case is the method. Change one element at a time, let the test run long enough to mean something, and decide in advance what counts as a win. The lesson from Upworthy is to watch what you optimize. A headline that wins on clicks but disappoints on the page is the clickbait problem again, and sometimes the right call is to pick the headline that is not the winner.

Small sites often lack the traffic for a clean headline test. Cheaper stand-ins are email subject lines (see our email marketing tips for 2026), two versions of a social post, or simply showing five drafts to five people who match your audience and asking what they expect to find behind each.

Headlines and search

A page headline and a search title are related but not identical. Google's documentation says to write descriptive, concise titles, give each page a unique one, and avoid vague labels. It also warns against stuffing keywords and repeating boilerplate. Google can also rewrite a title link when it decides the title does not match the page, so a clear, honest headline is also your best defense against a rewrite. NN/g suggests keeping page titles to about 40 to 60 characters. For answer engines, which reward pages that state things plainly, see our AEO checklist.

For the wider habit of saying plainly what you offer, see our timeless marketing tips that still work.

What to do this week

1. Pick your five most important pages or posts and read each headline out of context. Mark any that a stranger could not explain.
2. Rewrite each one three ways: one specific, one front-loaded, one with a number only if you truly have it.
3. Check each draft against the withholding and exaggerating definitions. Cut anything that fits either.
4. Read the first paragraph under each headline. If it does not pay off the promise, fix the paragraph or the headline.
5. Run one clean test on your highest-traffic page or your next email: one change, one metric, a fixed end date.
6. Log each winner and loser with the date.

FAQ

Is a curiosity gap the same as clickbait?

No. A curiosity gap becomes clickbait when the headline withholds the key fact or exaggerates and the page fails to deliver. If the answer arrives in the first few lines, you opened a gap and closed it.

Should I put numbers in every headline?

Only when the number is real and the page supports it. Specific detail helps, but an invented or inflated figure is the exaggeration Facebook described, and it costs you trust.

How many headline versions should I write?

Newsrooms in the Hagar and Diakopoulos study tested as many as a dozen per article, but you do not need that many. Write five to ten drafts, cut to two or three, and test only if you have enough traffic for a fair result.

Do I need an A/B testing tool?

Not to start. Subject-line tests in your email platform, two versions of a social post, or quick feedback from a few people in your audience will teach you something. Move to formal tests once you have steady traffic.