How does Google Discover work?

Google Discover is a feed of content offered on mobile, in the Google app and on the Chrome start page. Its peculiarity fits in one sentence: there's no query. You ask for nothing; the system suggests what it judges might interest you, based on what it knows of your activity. That's a complete reversal of search, and it changes everything — for the reader as for the publisher.

Discover, Search and Google News: three different things

What Google says publicly

The publisher documentation sets out a few explicit points, and we stick to those. Inclusion in Discover is not guaranteed, even for a perfectly compliant site. There is no application procedure: being indexed by Google is the condition, not a sign-up. A large, good-quality image makes content eligible for a more prominent display, without that amounting to a promise. And Discover's content policies are stricter than Search's on certain sensitive subjects.

Why its traffic is so unstable

An article can pull a considerable stream of readers through Discover in a single day, then fall back to zero the next, with nothing changed on the site. That isn't a malfunction: it's the normal behaviour of a personalised feed, where display depends on a particular user at a particular moment rather than on a stable ranking for a keyword. A publisher building its economics on Discover builds on sand — not through any malice on the platform's part, but through the nature of the product.

What Discover changes for the reader

The absence of a query shifts a responsibility. In search, you formulate a need and judge the answers against that need. In Discover there's nothing to judge against: the content arrives already chosen, and the only available criterion is whether you like it. It's an excellent machine for extending the interests you already have, and a very poor one for discovering what you don't know you're missing.

The counterweight is simple and fits in one move: keep at least one source you chose yourself and that reaches you with no intermediary — a subscription, a newsletter, an RSS feed. Not because Discover is harmful, but because an information diet made entirely of suggestions eventually stops surprising you, which is precisely the opposite of what information is for.

What we won't tell you

We won't describe the internal signals Discover supposedly uses to pick one article over another, because we don't know them and nobody outside Google does. Plenty of writing on the subject presents as fact what are really observations from a handful of cases and reasoning run backwards from traffic statistics. Those observations may be right; they remain hypotheses. We'd rather tell you honestly where the verifiable part stops.