Method / Indicator

Sourcing

What share of the article is attributed — a sentence that names who said it, quotes them, or credits a document. Points come off for "critics say" and the like, for contested labels with no one’s name on them, and for predictions nobody is quoted making.

0 unattributed named sources 100

How the score moves

Higher is better here. It starts from the share of sentences that say where what they assert came from: naming who said it, quoting them, or crediting a document
It then loses 12 for each claim credited to no one ("critics say"), 15 for each contested label used in the article’s own voice, 8 for each prediction nobody is quoted making, and 6 for each unattributed ranking ("the biggest day"), up to three
On our own corpus the middle article reads about 58; a heavily sourced report reads about 80

What is never counted

Quotations are not marked down — they are the attribution
Length: the penalties are a rate per ten sentences
A ranking that names who is making it

Highest in the last week

Lowest in the last week

Every reading is of the writing, not of whether the article is correct. The full method, including how the Niral Score is built, is on the Method page. Anything here can be checked against the API, which returns the same numbers with the words that produced them.

The other indicators

Niral Score Adjectives Repeated words Sentiment Political diction Generational Gender Jingoistic Religious Mood