How does Google News work?

Google News is an aggregator: it produces no articles of its own, it gathers those published by thousands of sites, clusters them by event and presents them as topics. Its promise is to show you several treatments of a given event rather than one. It's a useful service and, for many publishers, a traffic source — hence an abundant and frequently wrong literature about how to rank well in it.

The principle: clustering by event

The characteristic move of Google News isn't ranking, it's clustering. The system works out that forty articles from forty sites cover the same fact and gathers them into a single block with a representative headline and a list of coverage. That clustering is where the real value sits: it makes visible at a glance that a story is circulating everywhere — or, sometimes, that it circulates in exactly one place, which is itself a signal.

How a site gets in

For a long time you had to apply: a publisher submitted its site to a dedicated centre and waited for approval. That step was removed. Today any site is in principle eligible, and Google discovers news content through ordinary crawling, as it does for classic search. Eligible does not mean selected, though: discovery is automatic, display isn't.

What can honestly be said about ranking

Google publicly documents the broad families of criteria it says it takes into account — relevance to the topic, freshness, a site's standing on that subject, the user's location and language, plus their interests if they've expressed any. We stop there, deliberately. The internals aren't public, they change constantly, and everything written beyond those categories is conjecture — including when it's written with great confidence.

• There is no trick that guarantees appearing in Google News, and anyone selling you one is selling you a guess.
• No slot is owned: a site present today can be less visible tomorrow with no rule having changed for it.
• The service enforces published content policies, notably against deception about the publisher's identity and mass-produced content with no value of its own.

One practical consequence deserves stating: a publisher has no way of knowing why it's well or badly placed on any given day. It observes a result, not a cause. Any explanation it offered would be an after-the-fact reconstruction — and that's as true of us as of anyone else. We'd rather say so than let you believe someone, somewhere, holds the key.

Using it well as a reader

The best use of Google News isn't reading the top result, it's opening the cluster. Seeing ten different headlines on the same fact teaches more than one article does: the gaps between the phrasings reveal what's established and what's still interpretation. It is, paradoxically, an excellent verification tool — provided you use it against your natural reflex.