As of November 11, 2026, a shopping assistant is no longer a novelty at the edge of the funnel: on Douyin, chatbot links drove roughly 5% of October purchases, and more than 20% in considered categories such as supplements, according to WPIC Marketing + Technologies. That single number changes how a brand should read its own traffic. Referral share is now a line item, not a rounding error, and it comes with a gatekeeper attached.
From search box to answer box
Why a short list narrows discovery
The mechanics are easy to miss because they look like convenience. A shopper describes a need, the assistant returns three to five options, and the purchase happens inside the conversation. Compared with a page of search results, the short list compresses discovery into a single screen. For an established brand the effect can be neutral or positive; for a challenger that relied on ranking on page one, the list is a narrower door than the results page ever was.
Platforms have leaned in. JD.com said its Dongdong assistant drew nearly 80 million users for this year's festival, more than triple 2025. Tmall reported that AI token sales rose more than 700% between October 15 and November 5, and that AI toys passed 100 million yuan in sales within a year of launching as a category. Alibaba also said sales of AI companion robots rose nearly fivefold year on year in the second half of October. None of these numbers prove that AI lifts total demand, but they do show where platforms are steering attention.
The demand backdrop is thin
Syntun put Singles Day sales up 2.4% between October 7 and October 31, only slightly faster than the 1.1% rise in China's retail sales through end-September. Growth that modest explains the enthusiasm for anything that improves conversion inside an existing visit. When the pie is not growing, the contest moves to who controls the moment of choice, and an assistant that answers a question is closer to that moment than a results page.
The rules nobody published
The caveat comes from the analysts, not the platforms. Ben Cavender of China Market Research Group noted that some brands benefit more depending on how platforms are tuned and how they price, while others can be hurt if they no longer show up in product searches because shoppers let an assistant do the browsing. Jacob Cooke of WPIC called chatbot links a very new gateway and discovery channel, and expects them to take a growing share of direct purchases comparable to livestreaming. Both statements point the same way: the channel is real, and its rules are not published. Read as a market trend rather than a product feature, the shift mirrors what happened to livestreaming: a new discovery surface appears, category growth concentrates on the platforms that control it, and the brands that adapt their product pages and pricing first capture the demand.
What a brand can actually measure
Four readings survive scrutiny, and none of them requires guessing at the ranking model. Each one maps to a decision a brand can act on, which is the point: consumer demand is not expanding fast, so category growth has to come from conversion inside an existing visit.
| Metric | What it captures | Why it matters |
|---|---|---|
| Referral share | Orders tagged to assistant links | Sizes the channel before budgeting against it |
| Short-list presence | Whether the brand appears, and where | Being absent is different from being ranked low |
| Cited source | Pages the answer leans on | Shows whether discovery runs through owned or earned content |
| Refund rate | Assistant orders against search orders | Tests whether the channel brings fit or just volume |
The second row is the one most teams skip. A brand ranked third on a short list of four is still in the conversation; a brand absent from the list is not ranked at all. Treating both as a ranking problem leads to bidding for visibility that is not for sale, at least not publicly. The fourth row is the correction that keeps the first three honest: if assistant orders refund more often, referral share is buying volume rather than fit.
FAQ
Is chatbot referral traffic incremental?
Unclear, and probably mixed. Some shoppers would have bought anyway and simply asked first. The measurable version of the question is whether assistant orders carry a different basket size and refund rate than search orders on the same storefront.
Can a brand pay to appear in the short list?
Platforms have not published a paid placement product inside answers. What is documented is commission-type charging on completed transactions in some service categories. Treat paid influence as unverified and record labelling separately when it appears.
Why separate considered categories from impulse buys?
Because referral share diverges sharply. Supplements exceeded 20% on Douyin while the overall figure was about 5%, which means a blended average describes neither case and will misprice the channel.
What should be tracked weekly?
Short-list presence and cited sources. Referral share moves slowly enough for a monthly cadence, but the content an assistant leans on can change in days, and that is where a brand still has levers.
