What they did
Fishkin received leaked internal documentation for Google's Content API Warehouse from a source and reviewed it against Google's public statements. He published the first widely read account, describing how the code sat on GitHub from March to May 2024 before being spotted, and later named the source.
What they found
The docs contain Chrome-linked clickstream fields and host-level NavBoost signals. He quoted the documentation and pointed to a field named chrome_trans_clicks used for sitelink prominence, plus NavBoost data evaluated at the host level.
How much weight to give it
The documents show these fields exist and are named in the repository. They do not state how heavily any field counts, or whether it is live. Fishkin drew the inference that Chrome data reaching Search contradicts past Google denials.
What it doesn't prove
This is field existence, not proven ranking use. Fishkin explicitly warned readers not to point at a feature and call it proof Google uses it, since fields can be retired, test-only or internal. Google confirmed the leak was real but rejected the conclusions drawn from it.