AI News

AI Invades Double 11: Hands-On Tests of Taobao, JD AI Shopping, Doubao, and Zhang Dama Shopping Agents

36Kr · AIUpdated Nov 16 18:28

AI summary

Four AI shopping assistants—Taobao's built-in shopping Agent, JD AI Shopping, Doubao, and SMZDM's "Zhang Dama"—showed mixed performance in product selection, price comparison, and discount calculation during Double 11: Taobao's recommendations align with browsing history but offer few options, JD can provide historical price trends, Doubao can split orders across platforms to save money but easily loses context, and Zhang Dama falls short in linking multi-turn conversations. An Opinion Matters survey shows that 38% of consumers have used AI to make purchasing decisions, and 33% have bought products recommended by AI.

Body · AI translation

You can buy expensive things, but you must never overpay.

Everyone joins in the Singles' Day excitement year after year. By the time the final payment is due, people start regretting their impulse purchases.

Looking back at Singles' Day more than a decade ago, you really had to plan ahead.

Coupons, spending thresholds, cross-store order combining—you'd grab a piece of paper, fire up a calculator, and work out exactly how to buy things most cost-effectively. If you really couldn't reach the threshold, you'd have to rope in friends to combine orders.

Then you'd wait until midnight on November 11, place your order right on time, and sleep soundly.

Back then, shopping felt like solving a math problem, and snagging a bargain felt like handing in a perfect exam paper.

Today, the math problems are still there, but they keep getting longer, and consumers don't seem quite as fanatical anymore.

Presales, cross-store spending discounts, consumption vouchers, member discounts... Singles' Day kicks off early, and by the time the real November 11 arrives, the promotions have already been running for nearly a month. Not to mention the 618 before it, the Double 12 after it, New Year's Eve, and so on—there are so many shopping festivals that people have gradually lost the sense of occasion they used to feel.

Consumers no longer feel compelled to place orders at a specific moment. The promotions come around often enough—if you don't buy today, there'll be another one before long.

According to survey data from Opinion Matters, 38% of consumers have used AI to make purchasing decisions, and 33% have bought products recommended by AI.

It sounds like Singles' Day is finally about to transform from a mental workout into a relaxed, enjoyable shopping event.

To find out, we took the same shopping need and gave it to Taobao's built-in shopping Agent, the recently launched JD AI Shopping, Doubao—which can link to the Douyin Mall with one click—and "Aunt Zhang," the AI shopping assistant under SMZDM, to see whether they can actually sort out product selection, price comparison, and discount calculations.

01

They all look like they're helping you pick things, but the logic behind them is very different.

· Taobao: Before you even speak, it's digging up your shopping history

Taobao's AI assistant is embedded directly in the platform—open it and you get a chat box. Besides features like AI savings, order services, and a message assistant, it recently launched a dedicated "Singles' Day Shopping List" module, offering official shopping list entries across different themes such as popular must-haves, hidden-gem digital products, and wardrobe refreshes.

Click on a list and recommendations come up quickly. What's interesting is that these products aren't simply copied from trending rankings—they support personalized recommendations, bearing traces of brands the individual has browsed before.

Even if you don't go through the shopping list entry and instead just type "Give me a Singles' Day shopping list" into the chat box, the results will still lean toward products you've looked at before.

Especially on Singles' Day, when it's easy to browse and forget what you originally meant to buy, personalized recommendations really can save you some detours. For people who don't yet have a clear goal and just want to see what's worth buying, it's also easier to find shopping inspiration from familiar categories.

Of course, remembering search history doesn't equal truly understanding needs. Searching a certain category before might mean you were about to order, or it might just have been a casual look. Whether recommendations can further distinguish "interested" from "really want to buy" still depends on the subsequent product-matching capability.

But at least in the first round of interaction, Taobao holds a natural advantage: it doesn't just know what to talk about—it can connect the conversation with shopping behavior on the platform.

Although JD AI Shopping is a standalone app, after logging in with your phone number, favorites, cart, browsing history, and other existing information from the JD app are automatically synced over.

This means it's not a completely unfamiliar tool that knows nothing about the user.

However, when the same sentence "Give me a Singles' Day shopping list" was sent over, the result was somewhat different from Taobao's.

It gave a fairly common, generic shopping list, without clearly following previous search history. In other words, although it knows what information is in the account, it hasn't fully translated that into recommendation preferences.

Being able to read user data and being able to use user data well are not the same thing.

The former is about whether the information can be obtained; the latter is what determines whether recommendations fit the needs closely enough. For consumers, what's truly valuable isn't seeing the popular products everyone else can see again, but quickly finding the few that suit them best from a pile of products.

· Doubao: Set the rules first, then arrange the shopping list

Doubao, always eager to join the fun, also recently launched a "Ask Doubao on Double 11" feature.

Click in and you can first claim spending-discount coupons; after binding your Douyin account, you can use them directly to shop in the Douyin Mall.

Its feature design is fairly detailed too—there's picking products by budget, what to buy on Singles' Day, as well as product comparisons, ingredient and parameter analysis, and gift-giving guides, covering the shopping needs of different groups.

But what's really interesting about Doubao isn't how comprehensive its features are—it's that it seems especially "opinionated" when giving shopping advice.

When asked to recommend a Singles' Day shopping list, it didn't start by listing various popular products. Instead, it first laid down a few consumption principles: only stock up on daily necessities you'll use up within six months, only buy autumn/winter essentials and digital consumables, don't buy things that will sit unused, and don't force yourself to combine orders just to hit spending thresholds.

It also split the list into two parts: "must buy" and "don't buy." Tissues, large bottles of mineral water, thermal underwear, power banks, and charging cables were put on the must-buy list, while internet-famous ornaments, excess clothing, large packs of snacks, and useless small appliances were talked out of.

Doubao's Singles' Day shopping list

This attitude fits well with Doubao's consistently distinct, personality-driven style of expression.

Of course, this list is still skewed toward general advice and may not suit everyone's actual needs.

But in a scenario like Singles' Day, where it's easy to be led along by discounts, daring to clearly say NO is also a differentiated approach.

· Zhang Dama: Chat first, then pick products

Compared with the previous contenders that immediately present a dialog box, Zhang Dama's opening feels more like a lightweight interactive game.

A little girl first asks what you would like to be called, then learns your gender, birthday, and hobbies. After each answer, she gives corresponding feedback. For example, after you fill in your birthday, she responds: "I've remembered it. I'll send you a birthday wish on time."

This design feels a bit friendlier than a cold dialog box, and it makes information collection feel less like filling out a form. But cute is cute; what really determines the shopping experience is whether this information can be put to good use.

When the same shopping-list request is entered, Zhang Dama first asks which category of product you want to buy, then continues recommending based on the choice.

For example, if you choose digital products, its recommendations combine the brand of phone you currently use and offer related products from the same brand; if you choose daily necessities, the recommendations refer to the gender information filled in earlier.

Zhang Dama's Singles' Day shopping list

In this round of testing, it showed some ability to ask follow-up questions about needs and connect information, but the basis for recommendations was fairly limited.

In addition, Zhang Dama is connected to multiple e-commerce platforms such as JD.com, Pinduoduo, and Tmall, and can compare products and prices across platforms. Clicking a recommended product jumps to the corresponding platform.

This also makes its role closer to a shopping decision assistant rather than a guide for just one e-commerce platform.

Cross-platform price comparison is an advantage, but the shopping path is also fragmented as a result; consumers still need to handle ordering and order bundling themselves afterward.

Taobao's strength lies in its familiarity with shopping history, making recommendations more likely to align with existing preferences; JD.com has account information synchronization capability, but the lists it generates still lean toward generic recommendations; Doubao does not rush to list products, but first gives consumption principles, clarifying what is worth buying and what is not; Zhang Dama is more willing to proactively ask about needs and can also compare prices across platforms, but the depth of information collection and the continuity of subsequent purchasing still have room for improvement.

Behind this are actually four different shopping logics: AI embedded in platforms can use existing consumption data to shorten the decision path; independent AI shopping tools need to understand needs in greater detail; general-purpose AI starts from consumption advice and tries to help consumers reduce unnecessary spending; and the value of a cross-platform shopping assistant lies in breaking down the information boundaries of a single platform as much as possible.

02

When shopping on Singles' Day, the scariest thing is not that it is expensive, but that you spend a lot of money and then look back only to find it is cheaper elsewhere.

After all, being able to afford something is a skill, but buying it cost-effectively is real expertise.

Shopping needs: a skincare set suitable for sensitive skin,

an instant camera

, and a unisex branded sweatshirt, with a total budget of no more than 2,500 yuan. Taking into account cross-store full-reduction discounts, national subsidies, Singles' Day promotions, and other activities, provide the optimal shopping plan.

· Taobao: Not afraid of repeated changes, only afraid of having too few choices

In the first guide Taobao provided, it gave product combinations of three brands at different price tiers, with total prices of 1,150, 1,250, and 1,900 yuan respectively.

It also listed discount stacking plans, including suitable ordering time windows, cross-store full-reduction discounts, store coupons, 88VIP, and other discounts, while offering money-saving suggestions, such as using the saved budget to buy a few more packs of film.

This approach fits the consumer psychology of Singles' Day quite well: since an instant camera will always need photos taken after you buy it, rather than spending the entire budget on the camera, it is better to arrange the follow-up consumables as well.

Considering the upper budget limit of 2,500 yuan, this plan seems fairly restrained. Moreover, it recommended only one brand for each category, so the choice space was relatively limited.

When further asked whether the budget could be raised to 2,000-2,500 yuan, it immediately adjusted the product mix, switching the sweatshirt to a more expensive brand and giving corresponding reasons.

When subsequently asked about discounts, product details, and even requests to change products, it was basically able to keep up, with no obvious disconnect before and after.

It can be said that what really earns it extra points is not giving a perfect answer the first time, but being able to continue developing the entire plan after the budget and needs change.

From discount rules and the order of stacking red envelopes to product recommendations, the optimal plan, and ordering timing suggestions, it basically covered everything. The first shopping combination it gave was about 1,800 yuan, and it also indicated when during the platform's promotional period it would be more cost-effective to place orders.

In terms of product selection, the brands and products JD.com recommended for skincare and instant cameras overlapped with Taobao's; the difference was the choice of sweatshirt, and it provided two brand options for each category.

This arrangement has an advantage: rather than stuffing one product into each of the three categories and calling the task done, it leaves some room for substitution. When the budget is limited, adjustments can be made across categories, spending money on the products you care about more.

Subsequent interaction was also fairly smooth; changing products and continuing to ask follow-up questions basically stayed centered on the original needs.

In addition, JD.com can provide historical price trends for products, such as the low price over the past month and comparisons with recent prices. Taobao said in response that it "cannot read historical data."

This is especially important for consumers during Singles' Day. After all, whether a discount is substantial cannot be judged only by the crossed-out original price on the product page; you also need to see how much it actually sold for before.

Discount calculations solve "how to buy cheaply now," while historical prices help judge "whether buying now is really cheap."

Only by combining the two can it come closer to a shopping guide that can assist decision-making.

· Doubao: Can split orders to "hunt for bargains," but also has intermittent memory loss

The total price of the product combination Doubao gave the first time was lower, around 900 yuan. Budget control was quite cautious; you can tell it really tried hard to make users spend less.

Interestingly, besides recommending products for each of the three categories, it also provided a special ordering plan.

Doubao did not limit its recommendations to Douyin Mall products just because it was linked to a Douyin account; instead, based on subsidy levels and discount rules, it also took other platforms into consideration, and even said candidly things like "ordering on JD.com is cheaper."

It even proactively offered a "bargain-hunting" strategy, suggesting splitting the three categories into two separate orders to secure a better price.

The product combination guide Doubao provided, along with a split-order payment plan

After all, splitting into one more order could very likely mean using one more full-reduction discount and saving a few dozen yuan more.

When follow-up requests were made to raise the budget or switch brands, Doubao was also able to keep adjusting the plan. It would also provide related video guides on the Douyin platform as extra reference for understanding the discount rules.

However, over multiple rounds of conversation, a problem gradually emerged: requirements need to be stated specifically enough.

For example, when only a brand switch was requested without restating the budget range, Doubao might revert to the initial low budget and regenerate a guide.

It can offer order-splitting ideas and find discounts across platforms, but it does not yet consistently hold onto the context of the entire shopping task.

No matter how carefully the discounts are calculated, if detailed requirements get "lost" midway through the conversation, the final plan may still deviate from the goal.

· Zhang Dama: After the budget adjustment, the guide "went off track"

When it was Zhang Dama's turn, the style suddenly became "plain."

The first shopping plan it generated totaled only a bit over 800 yuan, and the recommended brands were all fairly budget-friendly.

It did save money, but the problem is that a budget cap of 2,500 yuan does not mean the money must all be spent, nor does it mean cheaper is always better. Whether the products meet the needs and whether the brands are suitable are equally important.

So the request was further made to raise the budget to 2,000–2,500 yuan.

Unexpectedly, Zhang Dama did not directly upgrade the product combination; instead, it began asking follow-up questions about who the purchase was for and the occasion, such as whether it was for oneself, family, or friends, and whether it was for a birthday, a promotion, or to lift one's mood.

I had assumed these questions would help make the recommendations more precise, but then it asked about the budget again, and what followed was still a single-product recommendation.

The budget and shopping needs that had already been stated earlier were not stably retained. After several rounds of conversation, the shopping plan that should have been gradually refined instead felt like starting a new game.

In the end, this guide neither fully answered how to combine the three categories of products nor provided a specific discount-stacking plan. As for how to shop during Double 11, it did not form a clear shopping path at all.

The only valuable thing might be... emotional value?

After all, there was always a cute little girl on the page, occasionally offering a line like "pleasing yourself is what matters most."

The product combination guide Zhang Dama provided; its multi-turn conversation ability is relatively weak

Taobao's multi-turn conversation was fairly coherent, but the AI recommended fewer product options; the JD guide was more complete with fuller pricing information, yet slightly verbose, raising the barrier to understanding; Doubao demonstrated cross-platform discount calculation and a unique order-splitting money-saving approach, but still has room to improve in consistently grasping context; although Zhang Dama proactively asks follow-up questions to refine needs, when facing complex tasks such as multi-category combinations and budget adjustments, the continuity of its multi-turn conversation still appears insufficient.

The real difficulty is whether product selection, budget allocation, discount calculation, historical prices, and the ordering path can be strung together into one complete plan. A single item being cheap does not mean the whole combination is the most cost-effective; listing many discount rules does not mean money was actually saved in the end.

03

AI starts "livestream selling," and GEO is about to take the table

No matter how polished the shopping guide is, if it is cumbersome to use, it is hard to keep people using it.

With relatively high public awareness and a low usage threshold, Doubao is easier for consumers to get started with.

By contrast, the other three products each make trade-offs in interaction design: Taobao is convenient and direct, but has only one chat window and does not support creating new conversation threads, so after chatting for a while it is easy to lose the key points; although JD AI Shopping can create new conversations to manage different shopping tasks separately, the jump links are slightly cumbersome; Zhang Dama feels more "human," but the deeper the conversation goes, the more its memory is tested.

Whether the budget has been fully used, whether the recommended products are suitable, and whether the so-called discounts are truly cost-effective still need to be checked by yourself.

Meanwhile, merchants are also beginning to face new problems.

GEO is therefore moving from content marketing to e-commerce.

Otherwise, no matter how detailed the earlier guide is, in the end you still have to flip through pages yourself and hunt for discounts, and the time saved gets spent right back.

Consumers ask AI to help them save money, merchants try to make AI see them, and platforms hope AI will keep shopping needs on their own turf. Double 11 is still the same Double 11, except that this time, what is being contested is not just prices and discounts, but also who can better control the shopping entrance in the AI era.

This article comes from the WeChat public account "Xiaguang AI Lab," author: Zimo, editor: Li Xiaotian, published by 36Kr with authorization.

Source

36Kr · AI