Rand Fishkin reviewed many US Congress documents from the hearing on "Online Platforms and Market Power: Investigating the Dominance of Amazon, Apple, Facebook, and Google" two days ago. He saw this document, which suggested "user signals like clicks" to Google users in Google search.
He posted this on Twitter:
SEOs will love this:
– Yes, Google uses "user signals like clicks".
– Yes, Google has some level of "domain" permission.
– Yes, they learn mechanically based on human assessment data
All these rejections all these years, but here it is, everything set out in the internal documents. / 4 pic.twitter.com/1ayo8mAGK7
– Rand Fishkin (@randfish), July 30, 2020
The line he refers to says, "Continuous investment in user signals like clicks. Our search users create the first level of the network effect of search quality, and we invest heavily in it."
No, where does that mean specifically that Google uses this as a "direct" ranking signal in the search. Google has indicated that it uses click data as an indirect signal. That said, Google will check if its algorithms work well and what to do by looking at user signals like clicks. But nowhere does that mean that Google attributes this directly to ranking algorithms. In fact, Google has repeatedly stated that it does not use this directly in rankings, but indirectly. Apple uses it and Bing does, but not Google – if you trust what Google has been saying for two decades.
Pedro Dias, who worked at Google some time ago, also summed it up:
Why is the fact that Google computers are learning against human data new?
– Pedro Dias (@pedrodias), July 31, 2020
Forum discussion on Twitter.