Introduction: what Search Quality Rating is

What the document says

The opening of the guidelines lays out Google’s worldview on search: people search for radically different reasons — from a quick task to deep research, from boredom to a critical life moment — and each kind of search demands a different kind of result. Medical results need to be trustworthy and aligned with expert consensus; a search for puppy photos needs to be adorable; a search for a specific website needs to deliver that site at the top. Ambiguous queries, or queries involving many perspectives, call for a diverse set of results.

The rater’s role is defined with a precision much of the industry ignores: no individual rating affects the ranking of any page. It would be unworkable — humans could never rate the open web page by page. The scores serve two purposes: measuring whether the search systems are working, and providing examples of helpful and unhelpful results that improve those systems.

The following sections cover the operational rules of the job: raters must represent the people of their locale (language + country), judging by local cultural standards and never by personal opinion, religious belief or political views; they must turn off ad blockers, because the page must be evaluated the way an ordinary user experiences it; and they must treat the document’s examples as the central calibration tool.

In practice

The guidelines are the specification of the target, not the mechanism of ranking. When Google ships a quality update — Helpful Content, core updates — the systems are being trained and measured against the standard this document describes. Optimizing for the guidelines isn’t “optimizing for raters”; it’s aligning your site with the definition of quality the algorithm pursues. It’s the most honest product document Google has ever published about search.

The concept of locale matters more than it seems. A Brazilian rater judges with Brazilian eyes: local cultural standards, expectations and references. Content machine-translated from English, examples that make no local sense, dollar prices on a .com.br site — all of it fails the test of representing the locale’s user. Localization is not translation.

Rating without an ad blocker is a message about experience. The rater sees your page with every ad loaded, like the ordinary user. If the real experience with ads is bad, that experience goes into the score — and that’s what the algorithm learns to demote.

The diversity of purposes protects niches. The document says explicitly that no page type is intrinsically superior: an encyclopedia is not worth more than humor. What gets evaluated is how well the page achieves its own purpose. That kills the idea that only “deep content” ranks — depth is a requirement of some purposes, not all of them.