The so-what

China just produced a clean natural experiment in what happens when a search engine's own AI ambitions eat its search-ad business faster than anything replaces it. Baidu's first-quarter 2026 results, reported May 18, show online marketing services revenue down roughly 22 percent year over year to RMB 12.6 billion, even as the company's AI-native marketing services line grew 36 percent to just RMB 2.3 billion 5 6. The gap between those two numbers is the story: the AI upside is real but still a rounding error next to what is being lost. Around that gap, a new industry is forming in real time, and one of its earliest players, PureblueAI, has just completed its second funding round in six months — a useful, fast-moving proxy for how institutional capital is pricing this transition.

The old model is bleeding faster than the new one is scaling

Baidu's headline framing was that AI-powered revenue now exceeds half of its general business for the first time, with core AI-powered revenue up 49 percent to RMB 13.6 billion and AI Cloud revenue reaching RMB 11.3 billion, infrastructure alone up 79 percent to RMB 8.8 billion 5 6. That is genuine growth. But it sits inside a company whose historic profit engine, search advertising, is shrinking on the back of Baidu's own decision to surface AI-generated answers across roughly 70 percent of mobile search results. Every query a large language model answers directly, instead of returning a ranked list of ten blue links, is a query that never reaches an advertiser's landing page. Baidu is not being disrupted by an outside challenger here; it is disrupting its own monetization model on purpose, betting that cloud and enterprise AI revenue eventually backfills the hole. The Q1 numbers say that bet has not yet closed the gap.

Why the traffic is actually moving

The scale of the underlying behavioral shift is what makes this more than a one-quarter accounting story. China's generative AI user base reached 602 million by the end of 2025, up 141.7 percent year over year, taking penetration to 42.8 percent of the population, according to the state-backed China Internet Network Information Center's latest statistical report 7. That is not a niche cohort experimenting with chatbots; it is approaching half the country routing everyday information-seeking through models like Doubao, DeepSeek, Ernie Bot and Kimi instead of a ranked results page. For any brand, retailer or service provider, the practical consequence is that the entire discipline built around ranking on page one of Baidu stops mattering once there increasingly is no page one — only a single synthesized answer that either names you or does not.

The new intermediary: generative engine optimization

That blank space is what generative engine optimization, or GEO, is built to fill: instead of optimizing for backlinks and keyword density, GEO tries to engineer a brand's odds of being the entity an AI model actually cites or recommends inside its answer. PureblueAI, a Beijing-based GEO specialist founded in 2025 by a team with Tsinghua and Chinese Academy of Sciences backgrounds and prior stints at ByteDance and Alibaba, has become an early bellwether for how fast that market is capitalizing. It first raised a seed round in September 2025 co-led by BlueFocus, a publicly traded marketing group (SZSE: 300058), alongside Inno Angel Fund 1 2 3. Six months later, in March 2026, it closed an angel round led by Vertex China, with Inno Angel Fund, 36Kr and Yicun Songling returning 4. Two rounds in six months, with the investor base widening from a strategic incumbent to a name-brand venture firm, is itself a signal: early believers are being joined by generalist capital willing to underwrite the category on its own terms, not just as a hedge.

BlueFocus's participation is the more revealing data point of the two. It marked the company's first AI-specific strategic investment since it declared an all-in-AI pivot in 2023, and it came from an incumbent marketing-services group whose own client base is exposed to exactly the disintermediation Baidu just reported 1 2. That is an agency effectively buying insurance against its own core business, rather than trying to build the capability internally — a pattern that tends to show up when the moat is judged to be technical (model architecture, citation-tracking data) rather than something an incumbent's existing distribution can replicate quickly. China's broader GEO services market had already reached an estimated RMB 167 billion by the end of 2024, equivalent to roughly 40 percent of the SEO market's size in the same period, according to Sina Finance's coverage of the BlueFocus deal 3 — meaning the category was already substantial before this latest financing wave.

What investors should actually track

The interesting number is not Baidu's aggregate claim that AI revenue now exceeds legacy ads. It is whether the AI-native marketing services line — the direct GEO-adjacent replacement for lost search-ad dollars, up 36 percent to RMB 2.3 billion — starts compounding fast enough to cross the shrinking RMB 12.6 billion legacy line within a handful of quarters, or stalls out as a permanent rounding error. Second, watch whether GEO spending shows up as a distinct, durable budget line in the earnings of China's ad-agency holding companies, starting with BlueFocus, rather than getting folded into generic AI-strategy commentary. Third, track whether PureblueAI's investor roster keeps widening beyond BlueFocus and Vertex China to include other strategics — telecom, e-commerce platforms, or additional agency holding groups — which would signal the market consolidating around a handful of category leaders rather than fragmenting across dozens of small GEO shops chasing the same 602 million users.