Thoughts on AI Search Products

Originally published in Chinese on February 1, 2024. What search is worth Search was the main traffic entry point of the last generation of the internet, and its business model is settled: it has supported companies worth tens of billions of dollars in country after country. For users it became the first door to the internet. As a product it is simple and covers almost everything. Because the form is so simple, doing it well is very hard, and the contest rewards whoever has accumulated the most technology and product skill. Attempts in the mobile era In the mobile internet era, many products tried to rebuild the value of search traffic by taking the search system apart and breaking the hold on the entry point. They tried different things at different times. Some tried to break the data barrier and rebuild search as a toolchain on mobile. Some tried to deliver value straight from search, which in China was called "search plus service". Others wanted to use search as a way in and become the user's operating center. Why the incumbents held Some of these strategies and products worked. But the search giants have deep technical foundations and do not get shaken easily, so the market stayed a duopoly or oligopoly and left startups little room. There is a second reason. The idea of search as the entry point has been drilled into users for so long that, whatever the problem, their first thought is to search for the answer. That mental habit built up over years, and time itself is a barrier. What decides the contest With that background, what would decide a contest in search? Where traditional search is strong Baidu and Google are very successful systems, and their results are among the best in the world. Each has its own core value. Google uses PageRank to judge how relevant one result is to another, and so surfaces what users more likely want. Baidu handles this in a similar way. The basic business logic of search is to trade search for advertising revenue. So the keys to a search product are experience, data, and user satisfaction, all of which look simple and are hard. Take response speed. Users have a limit on how long they will wait for each result, and if you take five seconds you lose that user. Accuracy and speed