What does a 5% share of orders actually measure? As of November 12, 2026, the clearest number from China's Singles Day is not the headline GMV but the share of purchases that start inside a chatbot. WPIC Marketing + Technologies estimates that AI drove roughly 5% of Douyin purchases through chatbot links in October, and more than 20% in considered categories such as supplements. JD.com says its assistant Dongdong drew nearly 80 million users during the festival, more than triple last year. For ecommerce teams, the interesting question is not whether AI converts, but what it converts and where the margin sits.
What the 5% actually measures
The 5% is a share of transactions, not of revenue, and that distinction matters. Chatbot-assisted baskets tend to be smaller and more research-heavy: a shopper asks which supplement suits a specific need, gets three candidates, then buys one. The category split confirms it — supplements, where comparison is the whole job, run above 20%, while impulse categories lag. AI is winning the part of the funnel where the buyer does not yet know what to buy.
That is also why the growth in assistant usage outpaces the growth in sales. Dongdong tripling to nearly 80 million users did not produce triple the revenue; it produced a larger pool of shoppers entering the comparison stage inside a platform app. For a brand, being absent from that stage is now a measurable loss, not a hypothetical one.
Two numbers that move in opposite directions
Singles Day sales rose 2.4% between October 7 and October 31 on Syntun's measure, against 1.1% retail sales growth in the first nine months of the year. Meanwhile cumulative sales passed 1 trillion yuan by October 31 — roughly US$154 billion — broadly comparable with the same period last year. An extended promotional calendar is holding the headline total up while underlying demand stays cautious, and AI tools are being used to defend conversion rather than to expand the basket.
Three entry points, three economics
| Entry point | What the shopper does | What the brand pays for |
|---|---|---|
| In-app AI assistant | Asks a question inside the marketplace app, gets a shortlist, buys in place | Listing quality and review depth; no separate media line |
| Chatbot link from a content app | Researches in a general assistant, then jumps to a product link | Commission on the resulting order, plus content that gets cited |
| Instant retail from a nearby store | Orders for delivery within the hour from a local node | Availability at the node, not just at the warehouse |
The three are not interchangeable. The in-app assistant rewards structured product data: if the listing does not state who the product is for, the assistant hedges or skips it. The chatbot route rewards citation — the assistant recommends what it can source and describe, which is why review depth and third-party coverage matter more than bid. Instant retail rewards physical availability, and it is the fastest-growing of the three: China's instant retail market crossed 1.2 trillion RMB in 2026, and Bain's read is that platforms are replacing traditional distributors as the gatekeepers to shoppers.
Budget follows the same logic. A brand that shifts spend from search ads into listing completeness may look like it is cutting marketing; in practice it is buying eligibility in two of the three entry points at once. Category growth in considered segments and consumer demand data showing which use cases buyers describe are the inputs worth collecting first. The third — instant retail — cannot be bought at all. It is won by having the item on the shelf of the node that serves the postcode.
Why instant retail changes the denominator
| Signal | Public figure | Implication for ecommerce teams |
|---|---|---|
| Instant retail market size, 2026 | Above 1.2 trillion RMB | An hour-level channel now sits inside the same demand pool as next-day |
| Platform investment in O2O | Over 100 billion RMB across Meituan, Alibaba and JD | Distribution is being rebuilt around nodes, not regions |
| Membership retail scale | Sam's Club China: 10.7 million paying members, about 140 billion RMB in sales | One listing can outperform dozens of provincial distributors |
When delivery promises drop from days to hours, the unit that has to be available changes from the SKU in a regional warehouse to the SKU on a specific shelf. That is a harder promise to keep, and it explains why availability, not traffic, is becoming the binding constraint in the fastest-growing channel. A campaign that drives demand into postcodes where the item is out of stock converts worse than no campaign at all.
What to fix before the next campaign
The practical response is unglamorous: measure fill rate by node, not by region; check whether the listing states the use case the assistant is being asked about; and treat third-party pages as part of your product data, because that is what the assistant quotes. None of it requires a new channel strategy. It requires accepting that the entry point moved, and that the content which used to sit below the fold is now the thing being read aloud.China Labour Bulletin: Singles Day tests AI shopping tools in China
FAQ
Is the 5% share a revenue figure?
No. It is a share of purchases completed through chatbot links on one platform in one month. Basket sizes differ from the platform average, so revenue share is likely lower than order share.
Why do considered categories over-index?
Because comparison is the task. Where buyers must weigh specifications or ingredients, an assistant that summarizes options replaces several searches, and the order follows the shortlist.
What should a brand fix first?
Listing completeness. Assistants answer from titles, bullets, reviews and Q&A; if the listing does not state who the product is for, the answer hedges or omits the product entirely.
How does instant retail fit in?
It is the availability layer of the same journey. Orders are fulfilled from a nearby node, so the constraint is shelf-level stock in the right postcode, not warehouse depth.
