RankBrain
Google's first deep-learning system for ranking, built to interpret the unfamiliar queries it sees for the first time every day and to adjust results based on how users respond.
Analogy Mental model
A manager reading the room. He sends a guest to table one, the guest recoils and asks to move to table three, and by tomorrow table three is at the front.
In practice What you actually do
Two behaviours matter here: whether your result gets clicked when it appears, and whether the visitor stays or bounces straight back to the results. Write titles that earn the click, then make sure the first screen delivers what the title promised.
Where people get burned Pitfalls
Bots that simulate clicks to game the system have been tried repeatedly and are detected quickly. The signal being optimised is genuine satisfaction, and a fake version of it is exactly what the countermeasures look for.
This entry was written in English from the Chinese original, not machine-translated. Read the Chinese version →