Bias spectrum
Haze maps source ratings across left, center, and right categories so readers can see whether coverage is clustered or balanced.
Rating methodology
How Haze rates sources and compares coverage.
Haze combines source ratings, coverage breadth, ownership context, and story-level distribution signals so readers can understand the evidence before deciding what to read next.
Factuality
Factuality labels summarize the reporting record, sourcing practices, corrections posture, and confidence we have in the available rating evidence.
Source selection
Source selection explains when Haze marks a publisher as original reporting, follow-up coverage, or contextual republication inside a story cluster.
Ownership
Ownership context shows whether a source is independent, corporate, public, state-affiliated, nonprofit, or otherwise institutionally influenced.
Blindspot
Blindspot signals highlight stories where one side of the political spectrum is covering an event much less than the other side.
Confidence
Confidence explains how strongly Haze matched a source, rating record, and story signal before showing that label in product.
Provider methodology
Attribution required
Step 1
Normalize provider labels into Haze buckets without hiding the original provider label.
Step 2
Attach provider attribution, version, effective date, and sync timing to every approved rating.
Step 3
Prefer normalized approved records and preserve uncertainty when provider records conflict.
Formula limitations
No single truth score
Source ratings describe publications, not whether one article is true or false.
Bias, factuality, ownership, and confidence are evidence signals, not a single weighted truth score.
Low coverage volume, stale provider data, or weak source matching keeps Haze from overstating certainty.