When an AI Agent Picks the Handbag, What Makes Yours the Answer?

Jing Daily published a piece on 5 October under the title “In the age of AI agents, luxury’s sea of sameness becomes a trap”. The argument is simple enough to state in one line, and I think it is the most under-discussed China marketing problem of the next two years.

When a person browses, sameness is survivable. You get seen anyway, because the shelf shows everyone. When a machine answers a question, sameness is fatal. The machine names two or three brands and stops.

I want to be careful here, because this is a subject where the available numbers are bad. Several firms publish Chinese AI search market shares. I have looked at them and I do not trust them enough to quote. What is not in dispute is the direction: Chinese users increasingly open an assistant, type a question in full sentences, and act on the answer.

Why luxury is unusually exposed

Read the China page of ten European fashion houses. Heritage since eighteen-something. Exceptional craftsmanship. Timeless design. Sustainable commitment. A founder with a vision.

Those sentences are not false. They are interchangeable, which is worse. A language model asked to recommend a French leather goods brand for a Shanghai buyer has to choose, and it chooses on the basis of what is specific, repeated across independent sources, and attached to a named thing.

Here is the mechanical consequence most brand teams have not worked through. The model is not reading your brand book. It is reading what other people wrote about you: Xiaohongshu notes, forum threads, media coverage, Baidu Baike, retailer descriptions. If what the market wrote about you is generic, you are generic to the machine, no matter how precise your internal positioning deck is.

Precedent is useful here. The same thing happened with search a decade ago. Brands that had written a real answer to a specific question kept the traffic. Brands that had written brochure copy lost it to anyone who bothered. Assistants compress that effect, because there is no page two.

What I am less sure about is the timeline. Assistants are not yet where most Chinese luxury purchase research happens. Xiaohongshu still handles more than 800 million searches a day and remains the centre of gravity. Anyone selling you an urgent AI pivot this quarter is ahead of the evidence.

What this means for your brand

The brands with the biggest exposure are mid-size houses with real distinctiveness that they have never written down in Chinese. You have a specific tannery, a specific town, a specific technique, and your Chinese content says “savoir-faire”. The machine cannot cite savoir-faire.

The brands least exposed are the ones with a fact so odd it is always repeated. A single workshop. A material nobody else uses. A founder story with a date and a place. Those travel through third-party text intact, which is exactly what an assistant needs.

An Italian accessories client of ours spent two years describing itself in Chinese as a maker of “premium Italian craftsmanship”. We changed nothing about the products. We rewrote the Chinese brand descriptions around one verifiable, specific fact about how the leather is finished and how long it takes. Within a few months the phrasing started appearing in notes written by people we had never worked with. That is the mechanism. Specific facts get repeated; adjectives do not.

What to do about it

Audit what the machine already says. Ask the Chinese assistants your customers use the questions your customers ask, in Chinese, and write down which brands get named. This takes an hour and most brands have never done it.

Find your citable fact. One thing that is true, specific, checkable and only yours. Not a value. A fact. If you cannot find one, that is the real finding and no amount of content will fix it.

Get it written by other people. Your own site matters least here. The fact needs to appear in Xiaohongshu notes, in Chinese media, in Baidu Baike, in retailer copy. That is a seeding and PR job, not a copywriting job.

Answer real questions in full sentences. Content built around the question a buyer actually types performs in both systems, classic search and assistants. We covered the search half of this in why search behaviour matters more than follower count.

On the generative side specifically, SEO Agency China’s work on DeepSeek and generative engine optimisation is the most useful English-language starting point I have found on how Chinese assistants assemble an answer and which sources they lean on.

Do not cut your Xiaohongshu budget to fund this. The social layer is where the citable facts get created. Starving it to chase assistants would be exactly backwards.

Where reasonable people disagree

One camp says this is the search transition all over again and the brands that move first will own the answers for years. Another says assistants will end up citing the same handful of authoritative sources regardless of what brands do, and the real work is getting into those sources.

I lean toward the second, which is why I keep pushing clients toward earned Chinese coverage rather than owned content. But I hold it loosely.

Have you asked a Chinese AI assistant to recommend a brand in your category? Did it name you, and if not, who did it name? Reply with what you found. I am collecting these.

Cindy Zhang, CEO of Fashion China Agency

Cindy Zhang is the CEO of Fashion China Agency. Fifteen years inside Chinese fashion and luxury: editorial and PR at Vogue China, brand-side experience with Guerlain, then the luxury segment at Xiaohongshu. She now leads a Shanghai-based team working exclusively with premium and luxury fashion brands in China. Read her full story →

Leave the first comment