Doubao and Qwen deployed full-scale AI commerce integrations during China's 618 festival, creating two distinct brand discovery architectures. Here's what it means for your GEO strategy.

China's 618 AI Commerce War: What It Means for Brand Visibility in 2026

China's biggest mid-year shopping festival just concluded on June 18, 2026 — and for the first time, the real battle wasn't over discount depth or traffic subsidies. It was over AI-assisted decision-making. Two platforms, ByteDance's Doubao and Alibaba's Qwen, deployed full-scale AI commerce integrations that fundamentally changed how Chinese consumers discover and choose brands. If your brand wasn't optimized for these AI surfaces, you likely missed a critical window.

The Scale of What Just Happened

The 618 festival ran May 15 to June 18, 2026, with three platforms — Taobao/Tmall, Douyin, and JD.com — running simultaneous month-long promotions for the first time in the event's history. But the more significant shift was structural: AI assistants became the first point of contact for millions of purchase decisions.

Doubao entered the festival with 345 million monthly active users (QuestMobile, Q1 2026), making it the largest AI assistant in China by consumer reach — ahead of Baidu ERNIE Bot (220M MAU), Quark (180M MAU), Yuanbao (150M MAU), and Kimi (90M MAU). Qwen, embedded inside the Taobao/Tmall ecosystem, claimed approximately 166 million MAU as of Q1 2026.

These are not marginal numbers. Together, the two platforms directly influenced purchase intent for hundreds of millions of consumers during one of the world's largest retail events.

Two Different Brand Discovery Architectures

Doubao and Qwen pursued different theories of brand discovery during 618 — and both matter for your GEO strategy.

Doubao's architecture: Content-to-Commerce

ByteDance's Doubao launched its full "Help You Choose" (帮你选) feature in May 2026, just before the festival began. The integration with Douyin created a closed-loop flow: a consumer watches a short video or live stream featuring a product, engages with Doubao to ask questions about it, receives AI-generated comparisons and recommendations, and completes the purchase — all without leaving the app.

This architecture means brand visibility on Doubao is partially upstream: it depends on how well your brand is represented in Douyin's content ecosystem. Brands with strong KOL coverage, high-quality live-stream presence, and positive user-generated content on Douyin had structural advantages in Doubao's recommendations. Brands relying primarily on standalone e-commerce listings faced a visibility gap.

Qwen's architecture: Intent-at-Search

Qwen's integration with Taobao/Tmall positioned it differently: as an AI layer on top of an existing purchase-intent moment. When a shopper opens Taobao and asks Qwen "which moisturizer is best for combination skin," Qwen generates a recommendation that surfaces specific brands and SKUs. Visibility here depends on structured product data, customer review signals, and sales velocity — traditional e-commerce signals, but now processed by a generative AI that determines which brands get named versus which disappear into a generic category answer.

The key tension CIW News captured in its 618 analysis: Qwen wants the assistant to remain visible as a brand advisor, while Doubao wants the transaction to feel effortless and invisible. For brand marketers, this distinction matters — Qwen's model gives brands more opportunity to be named and compared, while Doubao's model rewards brands that are already embedded in trusted content.

What the 618 Data Signals for Brand GEO

Several patterns emerged from the festival that have implications beyond the event itself:

1. "Decision-making power" replaced traffic as the primary competition axis

China's e-commerce intelligence community noted that 618 2026 marked a shift from "traffic competition" to "decision-making power competition." The platform that becomes the first AI touchpoint before a purchase — what CIW News called "the first words before placing an order" — exercises disproportionate influence over which brands get recommended. This is precisely the dynamic that GEO (Generative Engine Optimization) addresses: ensuring your brand is the one named when an AI answers a category-level question.

2. Doubao's advantage in lifestyle and consumer categories

Doubao's Douyin heritage gives it a structural edge in categories driven by visual appeal and influencer culture: beauty, apparel, food and beverage, home goods. In these categories, Doubao's content-first recommendation model means that brands with strong short-video presences are cited more frequently than brands with only text-based digital footprints.

In hubGEO's cross-platform scoring data, beauty and lifestyle brands consistently show divergent scores between Doubao and DeepSeek — with Doubao scores correlating more closely to Douyin content volume than to traditional SEO authority. This divergence is widening as Doubao's training data increasingly reflects its 345M-user behavioral graph.

3. Qwen's advantage in research-intensive categories

For higher-consideration categories — consumer electronics, automotive accessories, health supplements, B2B services — Qwen's intent-at-search model performs differently. Consumers in these categories arrive with more specific questions and expect structured comparisons. Qwen tends to generate more detailed, multi-brand comparative answers, which means there is more opportunity for brands to appear in citations — but also more risk of being unfavorably compared if your product data is not well-structured in Chinese.

4. The 618 citation gap for international brands

Foreign brands face a recurring disadvantage in Chinese AI recommendations that 618 made more visible: the absence of a Chinese-language entity layer. When a consumer asks Doubao "recommend a face cream for sensitive skin under 300 yuan," the AI draws on citations from Chinese review platforms (Xiaohongshu, Douyin comments, Zhihu discussions) and Taobao seller pages. International brands that have not invested in this Chinese-language citation infrastructure are systematically underrepresented — even if their products are available on Chinese e-commerce platforms.

During 618, this gap translated directly into missed discovery moments at scale.

The Citation Sources That Drove 618 Recommendations

Based on the observable outputs from Doubao and Qwen during the festival period, the following source types appeared most frequently in AI-generated purchase recommendations:

Xiaohongshu posts — especially those with high engagement and authentic user language, rather than obvious sponsored content
Douyin video metadata and comments — Doubao's training data includes rich signals from its parent platform
Taobao/Tmall product descriptions and Q&A sections — Qwen's primary structured data source for product-level recommendations
Zhihu answer threads — particularly for categories where consumers seek expert or peer-validated advice
Brand official accounts on Weibo and WeChat — as authority signals, though typically secondary to UGC sources

International brands that have localized their content presence across at least three of these five source types show measurably higher citation rates in hubGEO's scoring methodology. Brands present on only one or two — often just their official Tmall store — face significant citation disadvantages.

Takeaway for Brand Marketers

The 618 festival crystallized a shift that has been building for over a year: in China, brand discovery is increasingly mediated by AI before it is mediated by search, advertising, or even social feeds. Here is what the data from 618 suggests you should prioritize:

For brands targeting Doubao/Douyin audiences: Invest in Douyin content infrastructure — not paid advertising alone, but organic KOL content that generates authentic comments and community engagement. Doubao's recommendations during 618 tracked closely with a brand's Douyin content depth. If your brand cannot be cited from Douyin-native sources, Doubao will struggle to recommend you, regardless of your Tmall presence.

For brands targeting Qwen/Taobao audiences: Audit your Chinese product data. Qwen generates recommendations from structured signals: detailed product descriptions in Mandarin, well-answered Q&A sections, strong review volume with keyword-rich language. If your Tmall storefront was built primarily for human readers rather than AI retrieval, it may be underperforming as a citation source.

For all international brands: Run a cross-platform GEO audit before China's next major commerce event (Double 11 in November). Check your brand's scores on Doubao, Qwen, DeepSeek, and Kimi. If your Doubao and Qwen scores diverge significantly from each other, it is a signal that your content infrastructure is optimized for one platform's architecture and not the other. That asymmetry has direct revenue implications when the next 618-scale event arrives.

The brands that performed best in 618 AI discovery were not necessarily the biggest spenders or the most famous names. They were the brands that had built coherent, multi-platform Chinese-language presence — and the AI surfaces rewarded them for it.