Nearly 80% of consumers now ask AI before they buy
A new report out of Fudan University says the shopping funnel has quietly grown a new gatekeeper — and it answers in full sentences.
Nearly 80% of Chinese consumers now consult an AI tool before making a purchase decision, according to a national consumer spending-intentions survey for Q2 2026 cited in a new report on brand health in the AI era, released at a Fudan University seminar in Shanghai. The habits run through the entire buying journey: people ask chatbots such as ByteDance's Doubao and DeepSeek to compare options, weigh prices, pick a product, and even evaluate it after purchase. The report's framing is blunt — the competitive bar for brands has shifted from "being seen by users" to "being selected by AI."
The user numbers behind the habit are enormous. QuestMobile figures cited in the report put Doubao at 382 million monthly active users as of June, Alibaba's Qianwen app at 167 million, and DeepSeek at 130 million. "Asking an AI may gradually become a new entrance, and the path from recommendation to purchase is shortening," said Zhang Zhian, who heads the Fudan research center that produced the report. After two decades of brands fighting for position on a search results page, a single conversational answer is starting to sit between them and the customer.
The uncomfortable part is that most companies cannot see what that answer says. The report argues brand image is fragmenting into countless versions — different models, at different moments, in different contexts describe the same company differently, and firms control none of it: not the training data, not which sources get cited, not the wording. Ecosystem walls make it worse; as one executive quoted in the report notes, Doubao draws easily on Douyin and Toutiao content, while WeChat articles are mainly reachable through Tencent's own Yuanbao. Manipulation is also cheap: the report says "poisoning" software sells on e-commerce platforms for a few hundred yuan, and paid posts cost tens of yuan apiece at hundreds per day — enough to pollute how a model describes a brand within hours. Once negative information gets into a model's knowledge or retrieval range, it barely decays over time.
Why it matters: this is the consumer-facing version of the generative-engine-optimization scramble now hitting Western marketing budgets — except in China it has already reached mass adoption, with reputation risk moving, as one researcher quoted in the report puts it, "from public web pages into AI's cognition and dialogue." Whoever builds the equivalent of Search Console for chatbots — showing brands how often they're cited, ranked, and misdescribed — will own the next SEO-scale land rush. Until then, most brands are flying blind inside a black box that their customers trust.
What to watch: whether the major assistant makers open brand-visibility analytics APIs, and whether regulators treat paid manipulation of model answers as false advertising rather than marketing.
When was the last time you asked an AI before clicking "buy" — and did it change what you picked? Tell us in the comments.
