What happened
Google’s Content Warehouse API documentation was pushed to a public GitHub repository on 27 March 2024 and left there until 7 May. Erfan Azimi spotted it and passed it to Rand Fishkin of SparkToro, who verified it with former Google employees before he and Michael King of iPullRank published their analyses on 27 May 2024.
The documents describe 2,596 API modules with more than 14,000 attributes. Google later confirmed the documents were genuine, while cautioning against drawing conclusions from them out of context.
What it showed
The attribute names alone contradicted years of public messaging: measures built on click behaviour, a stored site-level authority value, per-document quality signals, and author information among them. For a community repeatedly told that clicks weren’t used in ranking, seeing click-derived fields documented internally was the story.
How much weight to give it
Moderate, and the caution is the point. These are real internal documents, which puts them well above speculation. They also carry no weights, no thresholds, and no confirmation that a given attribute reaches live ranking. A field can exist for experiments, for a deprecated system, or for a product other than Search.
What it doesn’t prove
Every confident claim of the form “the leak proves X is a ranking factor” overstates the evidence. The honest reading is narrower: Google measures more than it says, and its public denials were drawn more tightly than the systems behind them. For what’s actually confirmed about click signals, the stronger source is sworn testimony rather than this.