China's advertising incumbents just placed a bet that the businesses which decide what gets said about a brand to an AI chatbot will be worth owning outright. Two of the country's biggest marketing conglomerates, BlueFocus and Focus Media, each bought into a generative-engine-optimization (GEO) startup this year, racing to control how DeepSeek, Doubao and Qwen recommend consumer products. The catch: a ranking built to measure exactly that influence just showed it barely holds still from one edition to the next, and the acquirer with the bigger AI story to tell is still earning less than half a cent of profit on every yuan of revenue it books.

The land grab

BlueFocus, the Shenzhen-listed agency that runs China's core buying relationships with Meta, Google and TikTok, took a minority strategic stake in Beijing-based PureblueAI, co-leading the round with InnoAngel Fund alongside a Tsinghua alumni seed vehicle 3. The company has since told investors it has now backed six AI-native businesses spanning video generation, influencer-marketing automation, digital-employee infrastructure and ad-decision tools, with the PureblueAI deal marking its first dedicated push into GEO 5.

Focus Media, the elevator-screen advertising giant, moved with more conviction. In May it spent tens of millions of yuan to take a controlling stake in Titanimag, a GEO specialist that helped draft China's first industry standards for the category, pairing the startup's model-facing technology with Focus Media's offline ad-panel reach 4. Where BlueFocus bought a seat at the table, Focus Media bought the table.

Both deals point at the same emerging fact: China's GEO market was already worth an estimated 16.7 billion yuan by the end of 2024, equivalent to roughly 40% of the domestic SEO market, and it has been the fastest-growing line item in Chinese digital-marketing budgets since 3.

A ranking that will not sit still

The clearest evidence GEO is not yet a stable business came from the very firm the ad giants are chasing. On July 30, 36Kr and PureblueAI published the second edition of their Consumer Brand AI Recommendation Power roster, expanding coverage to five categories, phones, home appliances, autos, skincare and beauty, across 19 specific purchase-intent queries run against DeepSeek, Doubao, Qwen and Tencent's Yuanbao assistant 1. The scoring model weights a priority-recommendation-rate metric, essentially whether a brand is the first name an AI volunteers, at 40% of the total score 2.

Compared with the inaugural roster published in May, a meaningful share of brands that led specific purchase-intent categories lost ground by July, while others that were absent from the early rankings broke into the top tier 1 2. For an industry pitching itself as the new SEO, that is an uncomfortable admission: paid positioning in AI answers has not yet calcified into the kind of durable moat that made legacy search-engine optimization a decades-long business. It also explains why incumbents are buying rather than building. A ranking this volatile rewards whoever iterates fastest on prompt engineering and content structuring, which favors a startup over an agency's in-house team.

Thin margins behind the AI story

BlueFocus's own numbers complicate its GEO narrative. Full-year 2025 revenue reached 68.7 billion yuan, but net profit was only 225 million yuan, a margin of roughly 0.3%, because 97.9% of the company's costs are pass-through media spend paid straight to Meta, Google and TikTok on the overseas ad-buying business that still supplies more than 80% of revenue 5. Against that, AI-attributed revenue grew 210% to 3.7 billion yuan, but that is barely 5% of the total, not yet close to reshaping the group's economics 5. The stock trades near 157 times trailing earnings on the strength of the AI story rather than the AI cash flow 5.

The market has treated that gap as a trading opportunity rather than a warning. BlueFocus shares rose as much as 37% in a stretch beginning July 27 before giving back gains in volatile trading in early August, the kind of swing typically associated with speculative retail names rather than a low-margin ad agency 8. First-quarter 2026 results offered some real support: revenue up 31.9% year-on-year and R&D spending up 240% 6, and analysts tracked by domestic brokerages are now modeling first-half net profit anywhere from 146 million to 300 million yuan, a range wide enough to reflect genuine uncertainty about whether the AI pivot is translating into earnings or just funding it 6.

What investors should watch

BlueFocus reports first-half 2026 results on August 29, the first checkpoint that will show whether the AI and overseas dual-engine story management has been selling actually shows up as durable margin expansion rather than one more narrative rotation layered onto a pass-through media-buying business 6. Separately, watch whether PureblueAI or Titanimag publish a third-edition ranking before year-end: two consecutive reshuffles would confirm that GEO rankings behave more like paid search auctions, where position is rented every cycle, than like classic SEO, where authority compounds. That distinction determines whether BlueFocus and Focus Media just bought durable pricing power over China's AI-recommendation layer or a seat that has to be re-earned every quarter. With 602 million Chinese consumers now using generative AI tools regularly, a 42.8% penetration rate that grew 25 percentage points in a single year, the audience being fought over is not in question 7. Whether anyone can own the gatekeeping still is.