How it works

Last updated: 6 October 2026

Where an outlet's lean comes from

We do not rate outlets ourselves. Each lean is taken from published media-bias raters, mostly Media Bias/Fact Check, and also AllSides. Where they put an outlet on different sides, we leave it unrated rather than pick one. "Far left" and "far right" are counted with left and right. We keep a ledger of every outlet we have seen, with the rating applied and the rater it came from.

• No rating means unrated. If no rater covers an outlet, or the raters disagree, it is shown as unrated. We do not guess a lean from an outlet's name.
• Outlets with no politics have no lean. Sports, gaming, finance tickers and trade publications are left unrated on purpose.
• Estimates are marked and not counted. For some unrated outlets we estimate a lean from which stories they cover. These show as "· est." and are never counted in a story's left/center/right totals or in a blindspot.

Ratings are contested and no rater is authoritative. A lean describes an outlet's overall editorial position, not any single article.

State media

A left–right scale describes US politics. It does not describe a government's own news service. Outlets that a government, ruling party or state body owns or controls are labelled state media with their country. They count as coverage of a story, but on neither side. The test is ownership, which is public record. Publicly funded broadcasters whose charters guarantee editorial independence (the BBC, NHK, CBC, Voice of America) are not on this list.

What counts as one story

Headlines are grouped when they name the same people, places or organisations and share enough of their other words. A second pass joins groups that turn out to be one event reported in different words, such as one set of headlines naming a prize's winners and another naming only the prize. Each group needs a distinctive word in common, so separate events of the same kind (two unrelated police searches, for example) stay apart. The headline shown for a story is the one that best states what most outlets are reporting. It is not simply whichever arrived first.

Grouping is automatic and works on headlines only. It will sometimes split one event or join two, and a wrongly grouped story makes its coverage counts wrong too.

What a blindspot means

A blindspot says that outlets on one side covered a story and outlets on the other side did not. We only say so when all of these hold:

• the story is about politics, immigration, climate, crime and justice, or war and geopolitics. Sport, entertainment and science stories never get the flag;
• the story is at least three hours old. In the first hours, coverage reflects who publishes fastest;
• at least three outlets on the covering side reported it;
• the absence is unlikely to be chance. We do not follow equal numbers of outlets on each side. If a side makes up a small share of the outlets we follow, some stories will lack it by chance alone. So the more of the pool the missing side makes up, the fewer outlets it takes to call its absence a blindspot, and the less of the pool, the more it takes.

The outlet pool is not balanced

Most coverage reaches us through news aggregators and RSS. The outlets that publish most often are not evenly spread across the spectrum. On the day this page was written, the rated outlets on our front page split roughly 148 left, 137 center and 71 right, and over a third were unrated. Raw counts on a story therefore lean left before any outlet has made a choice. Read a story's split as "who among the outlets we follow covered this", not as "how the press as a whole covered it". Each story page therefore also gives every side's coverage as a share of that side's outlets, for example "reached 4% of the left-leaning outlets we follow and 6% of the right-leaning ones". That comparison does not depend on how many outlets each side has. The blindspot rule above is built on the same idea.