Baidu Still Owns China's Search. It Lost the Thing Search Was Worth.

Look at the scoreboard and Baidu appears to be winning. In June 2024 its flagship app had 703 million monthly users, up 4% year over year.³ It still answered something like 120 to 150 billion searches a month - six or seven times what ByteDance's Douyin handled.⁵ On mobile, where China actually lives, Baidu held 58% of all search at the end of 2024.⁴ By every metric you'd reach for to measure a search company, Baidu is fine. And yet everyone agrees it lost. The interesting question isn't whether the lead is gone. It's how a company can keep the customers and lose the business at the same time.

The official story is that Baidu fumbled the mobile transition - too slow, too desktop-bound, run over by the apps. That story is mostly wrong. The real story is that Baidu kept the searches and lost the searches that paid. Its core online marketing revenue fell 3% in FY2024 to RMB73bn even as its user base grew.² The erosion was financial long before it was visible in any market-share chart.

Not all queries are worth the same

A search engine is not paid by the search. It is paid by the auction behind the search. 'What's the capital of Mongolia' is worth nothing to an advertiser. 'Plumber near me,' 'best stroller 2024,' 'cheap flights to Sanya' - those are worth a fortune, because behind each one is a person about to spend money. Commercial intent is the entire asset. Baidu's whole machine was built to stand in front of that intent on the open web and rent the spot to whoever bid highest. So the right way to read Baidu's decline is not 'how many searches' but 'which ones.' And the high-value ones - local services, commerce, lifestyle - are precisely the categories quietly walking out the door into places Baidu cannot follow.

Two walled gardens Baidu's crawler can't see into

Here is the mechanism, worked down. A search engine monetizes the open web - pages it can crawl, index, and wrap in ads. China's most valuable content stopped being on the open web. WeChat launched in 2011, passed a billion users by 2018, and turned its Mini Programs - 4.1 million of them by 2023 - into an operating system for daily life: chat, payments, commerce, and the discovery that goes with them, all sealed inside one app.⁷ Then ByteDance did the same trick with attention: Douyin's short videos became where people now look up a restaurant, a product, a place to go. Analysts named the trap as early as 2019 - the large mobile apps are 'all closed system,' and Baidu's crawler simply cannot index what happens inside WeChat public accounts or Douyin videos.⁸ The query didn't disappear. It moved indoors, into rooms with no windows.

And once commerce moved inside, the payments moved inside too - which is the part Baidu never solved. WeChat Pay and Mini Programs let a user discover, decide, and pay without ever leaving the app. Baidu could send you toward a transaction; it could not own the transaction. That's the difference between a billboard and a checkout counter. ByteDance now captures roughly 21% of China's digital advertising market by sitting on shoppable attention,⁵ and most of that is money that, in the desktop era, would have flowed through Baidu's auction. The threat isn't even new: ByteDance has been circling search since Toutiao Search launched in August 2019 - the same year Baidu sued it in Beijing for allegedly scraping Baidu's own results - and Douyin Search in 2024 was already ByteDance's fifth attempt.⁶ Baidu saw it coming for half a decade. Seeing a structural force is not the same as being able to stop it.

Isn't this just a company that flubbed mobile?

The fair objection: this is the standard story of an incumbent too slow to make the jump, dressed up. It doesn't hold. Baidu made the jump. Its app grew to 703 million users³, and Baidu's own management has long described mobile as the majority of its search traffic.¹¹ If this were a distribution failure, the users would have left - and they didn't. The honest counter is sharper: Baidu's failure looks less like a company that missed mobile and more like Google failing to crack social - a perfectly distributed product that never built the second layer where the value migrated. For Google that layer was the social graph; for Baidu it was payments and commerce. Mobile distribution Baidu won. The app economy on top of mobile - the part that actually captures intent and closes the loop - it never built, and could not buy its way into against incumbents who got there first. Tellingly, even Tencent's effort to run a search rival, Sogou, ended with Tencent taking it fully private in 2021⁹ and absorbing it into its own ecosystem rather than running it as an independent challenger - Sogou Search still operates, but under Tencent¹⁰ - and it changed nothing, because the threat was never another search box. It was the disappearance of the open web underneath the one Baidu owned.

This is why the headline numbers mislead so completely. Baidu's total revenue in FY2024 was RMB133.1bn - down just 1% from RMB134.6bn the year before.¹ That looks like stability, even stagnation, not collapse. But peek under it: core ad revenue fell 3%, and the only thing holding the top line up was AI Cloud, where non-marketing revenue grew 18% in the fourth quarter.² The search business that built Baidu is quietly leaking its richest fuel; a different, newer business is patching the hole. A company can look flat on the page while its center of gravity moves out from under it.

Baidu was built to stand in front of the open web and charge admission to intent. It is still standing exactly where it always stood, doing exactly what it always did, with more users than ever. The world simply rebuilt itself around it - into gardens with no gates Baidu's crawler could enter, into feeds that end in a checkout it doesn't own. Baidu didn't lose the search box. It lost the web the search box pointed at. And you cannot win an auction for a transaction that now happens in a room you were never let into.