Appendices: the platform, duplicates and the changelog

What the document says

The platform (25.0–29.0). Raters work on raterhub.com, acquiring tasks that carry project-specific instructions (which override the guidelines). They can — and should — release tasks they can’t judge: the “lack expertise” examples include scientific, gaming and technical queries ([python pandas index like matrix]). The user’s location appears on a map with varying precision.

Duplicates (30.0). Two results are dupes when they have essentially the same content and the user wouldn’t benefit from seeing both. The definition is query-dependent: for a specific-content search (song lyrics), the same content on two sites is not a dupe — users want to verify accuracy across sources; for a broad query, pages with the same article and different ads are dupes — and the example used is the classic article replicated across article directories.

The changelog (Appendix 2) covers two years of revisions and maps Google’s recent priorities:

November 2023: the Needs Met scale simplified; examples modernized by “adding examples for popular platforms and video formats” — the official promotion of forums, social media and short video to first-class results.
February 2024: expanded characteristics of untrustworthy pages — the batch that brought fake author profiles and invented credentials.
January 2025: Lowest and Low realigned with the Web Spam Policies — expired domain abuse, site reputation abuse and scaled content abuse formally entering the raters’ manual, months after the 2024 spam updates.
September 2025: updated YMYL definitions and additional examples.

In practice

The changelog is the trail connecting the document to algorithm updates. The tactics punished by the March 2024 spam updates entered the human manual in January 2025; the November 2023 embrace of forums and UGC tracked the “hidden gems” era and Reddit’s SERP explosion. Reading the changelog at every revision is the cheapest way to anticipate where system training is pointing — and it’s why this guide will be updated with every new version of the document.

The dupe rule explains the death of syndication as a reach strategy. For broad queries, your article republished on another portal is a duplicate of the original — Google wants to show only one. Republication serves the host’s direct audience, not extra SERP presence; and without a canonical back to the origin, the “one” chosen may not be yours.

Even task releasing carries a message. The system acknowledges that judging technical content requires expertise — even a trained rater shouldn’t opine on what they don’t understand. Deep content in technical niches tends to be judged by people with more repertoire, which favors genuine depth over performed depth: the text that only looks technical impresses laypeople, not the person Google assigns to judge it.