How the Clickbait Score works

The score measures how hard a headline makes you work for the information it promises. It rates technique, never the topic or an outlet's politics.

Per article (0–100)

score = round(100 × (
    0.15 × promise_gap
  + 0.55 × curiosity_gap_headline
  + 0.20 × answer_position
  + 0.10 × min(ad_count / 30, 1)
))

promise_gap (0–1): how much the headline over-promises compared with what the article delivers. Rated by a language model.
curiosity_gap_headline (0–1): reliance on withheld information and bait patterns ("you won't believe", "what happened next"). Rated by a language model.
answer_position: how far into the article (by characters) the answer appears. Measured in code.
ad_count: ad slots detected in the page (ad iframes, ad containers, Google Publisher Tag slots, Taboola/Outbrain widgets). Measured in code.

For pasted text and screenshots there is no position or ad data, so only the two language-model terms are used, re-weighted to sum to 1.

Labels: low (<20), medium (20–39), high (40–59), extreme (60+).

Per site

A site's score is the average score of its articles summarized by BlinkRead users. Sites need at least 20 articles to be ranked. The weekly ranking uses articles from the last 7 days; the trend compares them with the previous 7 days.

Only public URLs count. Screenshots and pasted text are never published or included in the ranking.

Right of reply

Publishers can request a review of any article or score. We remove pages on request when the content is inaccurate.