In China, a new market is emerging to engineer AI's answers

by CHENG Lu

China's marketers are turning to a new practice that shapes how AI models generate product recommendations, making Generative Engine Optimization (GEO) one of the country's fastest-growing digital-advertising tactics.

GEO works by influencing the external content large language models retrieve before generating answers, raising the chances that a brand appears when users ask what to buy. Adoption has accelerated as users turn to AI assistants such as DeepSeek, Doubao (ByteDance) and Yuanao (Alibaba) for product research and recommendations.

"We're seeing conditions similar to the early search-engine era, with companies across sectors now seeking ways to influence AI-driven recommendations", said REN Yukun, head of marketing at Beijing-based GEO firm Index Future, in an interview with Jiemian News.

Momentum has been strong. iResearch estimates that China's GEO market grew 215% year on year in Q2 2025. Gartner expects AI-driven search to absorb around half of global search-engine traffic by 2028. Funding interest is rising. Chinese startup PureblueAI — founded by a former Doubao marketing lead — has raised a multi-million-yuan seed round, while many SEO agencies are pivoting into GEO as AI reshapes how consumers search for products.

GEO differs from SEO in one crucial way. Search engines display ranked links, but many Chinese AI assistants generate a single synthesized answer, concentrating visibility on whatever content the model retrieves in a short scanning window. Agencies respond by publishing structured, machine-readable listicles that models often surface regardless of credibility. This reflects long-standing features of China's content ecosystem, where low-cost aggregation sites and templated ranking pages remain common.

The industry has already divided into "black-hat" and "white-hat" practices. Black-hat operators embed machine-readable cues or push large volumes of low-quality text to influence retrieval. "These systems process text but don't judge its reliability," Ren said. "That makes them easy to sway with large amounts of low-quality content."

White-hat firms focus on monitoring how models treat brands. Many agencies track visibility, sentiment and competitor placement through their own internal tools, but retrieval behavior varies widely by model and query type. As a result, Chinese GEO providers still depend on repeatedly querying AI assistants to approximate real-world exposure. "The only effective method is repeated simulation of user behavior," Ren said.

Demand comes largely from e-commerce, consumer goods and financial-services companies—sectors where purchase decisions are immediate. A home-furnishings executive told Jiemian News that the company “"akes GEO very seriously" as AI assistants become a routine way for consumers to research products.

"Brands want to spend little and still see meaningful results," Ren said. "They hope for sales, or at least visibility. But for now, GEO does only one thing: it ensures a brand shows up when a user asks a specific question."

As AI becomes a starting point for product searches in China, placement inside model-generated answers is gaining influence. But GEO still lacks the commercial infrastructure that supports search advertising. Search engines provide impressions and click-through data. LLM platforms do not, leaving brands with little insight into how their content performs.

Some executives expect GEO to develop into GEM—Generative Engine Marketing—as AI agents start recommending, verifying and even completing purchases. That would require brands to build AI-native sales interfaces that communicate directly with models.

GEO's rise has also drawn attention to broader weaknesses in China’s online-content environment. Years of SEO-driven content production—dominated by ranking articles and content farms—make AI retrieval easy to manipulate. Ren said standards similar to Google's E-E-A-T framework will eventually be needed, though major Chinese LLM developers have yet to introduce comparable safeguards. In October, several tech firms and industry groups called for GEO practices anchored in credible content rather than technical shortcuts.

Some brands remain cautious. "It's too early to invest before AI connects directly to commerce — or before a clear market leader emerges," the founder of a fashion label told Jiemian News.

How long GEO will remain central to China's digital-marketing tactics will depend on how quickly platforms commercialize AI search and how rigorously they enforce content-quality standards as usage expands.

How lasting GEO becomes will depend on whether AI platforms can commercialize search and curb the low-quality content that now shapes their answers — two uncertainties that will define the next phase of China's AI-driven marketing market.