Gary Illyes from Google has provided a straightforward explanation of how the RankBrain algorithm functions. He addressed certain theories and misconceptions about the algorithm that have circulated.
Illyes is gearing up for a Reddit AMA session, scheduled for tomorrow (Friday) from 1 pm to 3 pm EST. The thread is already filling up with questions, and Illyes decided to answer one of these questions a day early.
The question was about RankBrain, specifically addressing claims that the system includes UX signals such as Dwell Time, Bounce Rate, and Click Through Rate. The inquiry also asked for clarification on whether RankBrain uses UX signals of any kind, in the context of serving results for unrecognized searches.
In his response, Illyes explained that RankBrain uses search data to predict what a user might click on when encountering a query Google has not seen before. He described RankBrain as a machine learning component that uses historical search data to anticipate the most likely user clicks for previously unseen queries.
Illyes further explained that RankBrain often addresses challenges that Google previously faced with traditional algorithms. RankBrain depends more on data from user interactions with search results, rather than interactions with content itself. He emphasized that RankBrain relies on historical data from results pages rather than the landing pages.
Finally, Illyes criticized those who complicate the understanding of RankBrain by suggesting it uses on-page signals, asserting that concepts like dwell time, CTR, and other theories are often unfounded, and that search is easier to comprehend than many assume.
If you have any other questions for Gary Illyes, make sure to submit them before Friday at 1 pm EST.
