Test bench · version 1
Media Rank
A ranking of the world's news outlets, assembled by machine from what publishers put out themselves: their feeds, the links the open web gives them, and how widely their name is known. No panels, no surveys, no editorial judgement.
This measures observable behaviour on the network, not truthfulness. A high score means an outlet is widely linked, publishes steadily, signs its work and is known by name — not that it is right. The originality block (who publishes first and who echoes) is not in this version yet.
| # | Outlet | Type | Country | Lang | Wiki readers | Langs | Rest | Score | Conf. |
|---|
By country
Each country keeps its own table: a national broadsheet and a trade weekly are not measured against CNN, they are measured against their own field.
By type
Newspaper, magazine, radio, television, online, agency — the type comes from the catalogue, so it is known for part of the corpus only.
How the number is made
Every figure is measured by us or taken from an open file. Nothing is self-reported by the outlet, and nothing is bought from a traffic panel. The ranking is recounted once a month — a place that changed every night would mean nothing.
Two tiers, and the tier is absolute
An outlet with a Wikipedia article always stands above one without. That is a decision, not a calculation: an encyclopedia article about a news outlet is the one piece of recognition nobody can buy, and editors in many countries had to agree it was worth writing. Within the upper tier the order comes from how many people read those articles last month, summed across every language edition. Everything else we measure only moves an outlet around that place.
Liveness moves an outlet inside its tier, never across the line: a silent outlet with an article sinks to the bottom of tier A, but does not fall under a busy outlet without one.
Network weight — who links to you
Ahrefs Domain Rating, harmonic centrality in the Common Crawl host graph, and the number of Wikipedia language editions carrying an article about the outlet — not SEO links, but what editors around the world judged worth an encyclopedia entry.
Output — how much and how evenly
Items per day (median, logarithmic), regularity — the share of days with at least one item — and how many hours of the day the newsroom is active. A feed that fires everything at midnight looks different from one that works through the day.
Newsroom signals — is there a newsroom behind it
How many distinct human bylines appear in a month, how many sections the feed carries, and what share of items carries an agency credit instead of a reporter. A feed with no bylines and no sections scores nothing here — absence of evidence is not cleanliness.
Reach — how far the name travels
The Tranco popularity bucket and how many people opened the outlet's Wikipedia article last month, summed across language editions. We have no branded-search data and never will, so encyclopedia readership stands in for it. It measures how well the name is known, not how much traffic the outlet gets: The Guardian drew 76,895 readers to its article last month, Chalkbeat 477.
What the score does not do
- It does not measure accuracy, bias or trustworthiness. Those are judgements, and a crawler cannot make them.
- It does not yet separate an originator from a republisher. That needs story clustering across outlets, and it is the next block.
- An outlet that blocks our crawler is missing data, not a zero — it is absent from the snapshot rather than punished.
- Confidence is published next to every score. An outlet found yesterday and an outlet watched for a month cannot be judged with equal certainty.
- Country comes from the national domain zone where there is one; for
.comand the like it comes from the catalogue, which is sometimes wrong. Those are marked with a dashed border.
Domain Rating by Ahrefs. Host graph by Common Crawl. Popularity bucket by Tranco. Article counts and readership from Wikidata and Wikimedia. Feeds are read from the publishers themselves, under the rules in our crawler policy.