For the first time in the history of China's biggest online shopping festival, four AI shopping assistants entered the 618 battlefield simultaneously. Doubao from ByteDance. Qianwen from Alibaba. JD AI Gou from JD.com. And Xiaohongshu's Wenwen. Each brought their own approach to what could be the biggest structural shift in e-commerce since mobile.
Here is what happened, who is winning, and what it means for brands advertising in China.
The 618 2026 AI shopping assistant war is not about which model performs better on benchmarks. It is about which platform owns the data pipeline from content creation to transaction completion. Doubao has already closed that loop. Xiaohongshu is the only credible challenger. The rest are playing a different game.
📊 Who Is Competing
| Platform | AI Assistant | Parent Ecosystem | Go-Live Date | Data Strategy |
|---|---|---|---|---|
| ByteDance | Doubao | Douyin + TikTok Shop | 618 test, Q4 full launch | UGC video + transaction = closed |
| Alibaba | Qianwen | Taobao / Tmall | May 11, 2026 | Transaction only, no content data |
| JD.com | JD AI Gou | JD.com | Dec 2025 app, 618 expanded test | Product SKU data, no UGC |
| Xiaohongshu | Wenwen | Xiaohongshu | Dots AI dept announced Apr 2026 | Deepest UGC, weakest transaction |
Alibaba struck first on May 11 by connecting Qianwen directly to Taobao. JD.com had already launched its standalone JD AI Gou app in December 2025 and expanded testing for 618. Xiaohongshu announced a new Dots AI department in April 2026. ByteDance waited until 618 to reveal its hand — and what a hand it turned out to be.
🔀 Two Approaches: Discovery vs Search
Hands-on testing by multiple Chinese media outlets during 618 revealed a clear split in product philosophy:
Route A: Conversational Discovery (Doubao, Qianwen, Xiaohongshu)
These assistants treat shopping as a conversation. They ask follow-up questions. They narrow options step by step. Each response ends with a question to refine the recommendation. When you ask for a pair of headphones under 500 RMB, they don't just dump a list — they ask what type, for what use case, with what priorities.
Route B: Wrapped Shelf Search (JD AI Gou)
JD AI Gou takes a fundamentally different approach. When asked for headphones under 500 RMB, it returned 15 products across 5 categories in a single response. No follow-up questions. No pros and cons. The interface essentially re-routes the existing product search engine through a chat window. Under the hood, the keyword still matters more than the conversation.
JD AI Gou's accuracy issues were serious. Testing by Sina Finance found the assistant claimed two Edifier ear-clip headphones used bone conduction technology — they don't. It recommended the same product (Moto Buds Clip) in two different categories from two different store listings. These are not UX imperfections. They are structural failures that expose JD's reliance on SKU search logic rather than conversational reasoning.
🔒 Why Doubao Is Winning: The Closed-Loop Advantage
The dominant takeaway from 618 across all independent reviews is clear: model quality will converge within 12 months. The structural advantage belongs to whichever platform owns the content-to-transaction data pipeline.
Doubao is the only platform that has already closed this loop.
UGC video content (product reviews/unboxing/experiences) from Douyin feeds directly into training data. Transaction links from Douyin Shop connect content to purchase. When Doubao recommends a product, it references real user experiences.
China's deepest consumer review database with exceptional product coverage in beauty, fashion, home, and outdoor categories. Transaction volume is a fraction of Douyin Shop. Content depth is the advantage — if transaction data catches up within 12-18 months, Xiaohongshu becomes the only credible challenger.
in 618
~$0.17
Ad Spend 2026
vs 2025
with AI Shopping Ads
AI Search Tools
Content-Transaction Loop
Model Parity
💥 What This Means for Advertisers
1. From Keyword Bidding to Conversational Intent Capture
Doubao and Qianwen are shifting purchasing decisions from shelf browsing into AI conversations. When a user asks an AI assistant "what laptop should I buy under 3,000 RMB," the recommendation pipeline is fundamentally different from keyword search. Brands need to understand how conversational AI ranks products — and it is not the same as search ad ranking.
2. Content Assets Become Ad Inventory
In Doubao's closed-loop model, authentic user reviews, unboxing videos, and product experience content directly influence AI recommendations. This means every piece of UGC content a brand generates on Douyin is now a potential ad asset feeding into AI recommendation systems. This is a structural change, not a marketing trend.
3. The Trust-Accuracy-Attribution Triangle
JD AI Gou's accuracy problems expose a critical risk: when an AI shopping assistant recommends the wrong product or misattributes features, consumer trust collapses. For advertisers, this means platform selection must now consider recommendation accuracy alongside traditional metrics like reach and CPM. A high-reach platform with inaccurate recommendations is worse than a smaller platform that gets it right.
4. Douyin/Doubao Deserves Priority Budget Allocation
As the only platform with a closed content-to-transaction data loop, ByteDance's Doubao currently holds the strongest structural position. Q3 2026 will see Doubao integrate with Douyin Shop subsidies. Q4 will see full commercial rollout. Brands that test conversational commerce formats now will have a head start when the platform flips the switch.
5. Xiaohongshu Is the Long Bet
For beauty, fashion, home, and outdoor brands, Xiaohongshu's Wenwen could become a high-conversion AI shopping assistant within 12-18 months if it closes its transaction gap. Brands should maintain content investment on the platform now, treating the current period as training data accumulation for future AI recommendations.
🚀 What You Can Do Now
1. Audit your Douyin UGC content pipeline. Are real users generating product reviews, unboxing, and experience content? This content will feed AI recommendation training data.
2. Test Doubao shopping queries for your category. Ask Doubao "what [product in your category] should I buy" and see what it recommends. If your brand does not appear, you have a content gap.
3. Allocate 5-10% of Q3/Q4 Douyin budget to conversational commerce testing. Don't wait until the platform commercializes the feature. Be ready.
4. Continue Xiaohongshu content investment. The platform's UGC depth is its structural advantage. Brands that build content now will be rewarded when Wenwen's transaction integration matures.
Don't put all your 618 budget into one platform. Split across Baidu (40%), Douyin (35%), and Xiaohongshu (25%) for maximum reach. Use platform-specific AI tools rather than one-size-fits-all creative.
🎯 The Bottom Line
618 2026 will not be remembered as the year AI took over e-commerce. It will be remembered as the year we saw the blueprint. Doubao's closed-loop system revealed what AI-native commerce actually looks like: a conversation, not a search result. Content that was previously "awareness" is now "training data." And the platforms that own the full pipeline from user experience to transaction will own the next decade of commerce.
CIW News (June 4, 2026), Sina Finance (June 3, 2026), 36Kr (May 11, 2026). Hands-on testing results compiled from multiple independent Chinese media outlets during 618.