er Googler: Google uses clicks to rank rankings //

“Almost everyone knows that we use clicks to rank websites. This is the debate: “Why are you trying so hard to hide this issue when everyone knows?” ‘”

This quote is from Eric Lehman. He was a 17-year Google employee who worked on search ranking and quality. He left Google in November.

Lehman gave testimony last Wednesday in the ongoing U.S. antitrust trial against Google.

You will hear this quote a lot if you haven’t already. You’ll hear it a lot.

But. This is not all Lehman said. He said that Google’s BERT and MUM machine-learning systems are more important than the data of users.

Fortune reported that Lehman believed Google would rely more on machine learning for evaluating text than on user data.

Training data vs. user data. BERT also caused confusion about “user data”. Big Tech on Trial published:

The DOJ’s attempts to impeach Lehman also appeared to backfire. Lehman responded to a DOJ inquiry about whether Google has an advantage over competitors in using BERT because of its data. He said that Google’s “biggest advantage” in using BERT is that it invented BERT. DOJ put up a bullet point exhibit for Sundar titled “Bullet points”. One of the bullet points on the exhibit stated the following: Our training data gives a significant advantage. By using BERT to fully utilize the training data and serving it to users …’, we can maintain and extend the lead.

If “training data” was a type of user data, this would have likely been a good impeachment against Lehman. After DOJ finished its re-direct exam, Judge Mehta questioned Lehman about “training data”. Lehman clarified that it was not the same as user search data.


How does it feel to compete with Google?

Sensitive Topics. DOJ attorney Erin Murdock Park also asked Lehman about a slide in one of his “Sensitive Topics’ slide decks that instructed employees not to “discuss the use of Clicks in Search …”

Lehman, according to a report from Big Tech on Trial via X, said that “we try not to confirm that we use the user data to rank search results.”

The reporter said: “I did not get good notes on this. But I think it had something to do because they didn’t want people to think SEO could be used as a way to manipulate search results.”

Google = Liars? SEOs have used Lehman’s quotes to prove that Google has lied about click-through rates or clicks for the past 25 years.

Last week, during an A.M.A. with Google’s Gary Ilyes in Austin, the question was asked whether Google uses clicks. Illyes’ answer was, “technically yes” because Google uses historical data to train its machine-learning algorithm RankBrain.

Translated from Googlerspeak, yes is technically true. RankBrain is trained using user search data.

Illyes told us about this in 2018. He stated that RankBrain uses historical search data in order to predict which query a user is most likely to click on.

RankBrain is used for all searches, impacting a “lot” of them starting in 2016.

Google Search tracks all clicks. However, the fact that Google Tracks Clicks in Search doesn’t mean they are used to rank directly. If site A receives 100 clicks while site B receives 101 clicks, site B will automatically move up to position 1.

Google’s ranking system is likely to be trained by clicking on the results of queries, much like the way it uses people to rate its search results.

Why do we care? Does Google track clicks on its website? Yes. Again, I’m not sure if it’s a ranking factor (though I’m not a Google employee and don’t have access to their algorithm). I’m aware that clicks can be noisy and easily manipulated. Google would not be able to use this ranking signal as it is simply too small for most sites and queries.

Dig deeper. Google’s algorithm is a mystery: everything you ever heard about clicks, bounce rates and CTR.

The original Search Engine land article Former Googler : Google uses clicks to rank rankings was first published.

Leave a Reply

Your email address will not be published. Required fields are marked *