<b>The 2024 Content Warehouse leak: read it as vocabulary, not as a manual</b>

<b>The 2024 Content Warehouse leak: read it as vocabulary, not as a manual</b>

The question: what did the May 2024 leak of internal Google API documentation actually tell us about trust signals?

The leaked Content Warehouse documentation exposed thousands of attribute names from an internal data store. Several are suggestive for E-E-A-T discussion: fields referencing site authority concepts, host-level signals, author identification, and a 'siteAuthority' attribute among them. Commentators seized on these as proof of a domain-authority-like metric.

The disciplined reading is narrower. The leak shows that attributes <i>exist</i> in a storage schema. It does not show whether they are used in ranking, how they are weighted, whether they are deprecated, or whether they feed live systems at all. A field name is a label on a box, not evidence of what is in the box or whether the box is opened.

What it does usefully confirm is that Google's internal model of the web includes host-level and author-level concepts — consistent with the idea that some signals operate at site and entity scope, not only per-page. That aligns with the reputation-research framing in the QRG.

Caveat: Google confirmed the documents were genuine but cautioned against drawing conclusions about ranking from a storage schema. That caution is methodologically correct, not merely defensive. Inferring weights from field names is a category error.

What we still don't know: which of the leaked attributes are live, their weights, and whether the trust-relevant fields feed ranking or some unrelated internal process.