On 9 November, Google launched a Chinese-language shopping feature for Gemini in Taiwan. Customers only have to describe what they need, as if they were chatting with a friend, and Gemini recommends a few products straight away, lays them out in a carousel of cards with product images and links to buy, lists prices, reviews, stock and where to buy, and adds a spec comparison table. That means AI picking products for shoppers in Taiwan has been live here since that day. Which products and which shops a customer sees in Gemini is being decided in every conversation right now, and anything that isn’t picked doesn’t even get compared this time round. This step happens before the customer pays, and it has nothing to do with whether people can pay directly inside Gemini, so the time to act is now.

What happened

The news comes from a 9 November announcement on Google’s official Taiwan blog, and Taiwan’s Central News Agency reported it the same day. The examples in the announcement are everyday ones: picking a dehumidifier for a studio flat in Taipei, choosing a carry-on that’s easy to wheel before a trip abroad, finding a present for a year-end gift exchange. Customers spell out their budget and what matters to them, and they can ask Gemini to pick out a few and compare specs, prices and reviews side by side.

The feature is backed by the Google Shopping Graph, a database of products. Google says that it holds more than 50 billion product listings, and that as many as 2 billion records in it are updated every hour.

The payment side needs to be clear from the start: what the announcement describes is links to buy on the cards, and it doesn’t mention paying within the conversation. The Taiwanese tech site KoCPC also wrote in its report the same day that Google’s post this time gave no timeline for when features such as price tracking and automatic checkout would come to Taiwan as well. Gemini’s help page does already list Taiwan among the places where its shopping features are supported, and the same page describes tapping “Buy” in the conversation and checking out with Google Pay, but it doesn’t say in which countries that button is available.

Why it matters to you

Whichever shop a customer ends up buying from, the shortlist is already set before they click a link to buy. For Gemini to put your product in the carousel, it at least has to be able to find you in the information that it can read, such as the Shopping Graph mentioned above; as for how it picks those products from there and what it looks at when ranking them, Google’s announcement doesn’t say. And once you’re in, the price, reviews and stock for your product in Gemini’s answer are based on what it read: if what it read is an old price, an old price is what the customer sees.

For example, one of the sample requests that Google gives reads, in translation: “I’d like to pick an exchange gift for a friend who loves camping and pour-over coffee. The budget is NT$1,500. Please recommend 5 quality items and where to buy them.” Suppose you run a small online coffee shop in Gukeng, Yunlin, roasting your own beans, and you sell a pour-over starter set that’s easy to take camping, priced within that NT$1,500; that request is practically asking about your product. Once a customer asks it, Gemini recommends five items as asked, and whether you’re among them, and how your price and reviews are described, all gets laid out on the customer’s own phone or computer, where you’re not there to see it.

What to do about it

You need to act, because Gemini has been drawing up these recommendations since the day the feature launched. Every time a customer asks a shopping question, it draws up another set, and every time you’re not in it, someone else is. The very start of Google’s announcement is about choosing presents for family and friends, and it names the year-end shopping season too. That season won’t wait for you to be ready.

There are two reasons you might not show up in the recommendations. One is that your product genuinely doesn’t meet the conditions that the customer gave, say it’s over their budget, and then it’s only natural that it isn’t there. The other is that it does meet the conditions but still doesn’t get recommended, and that’s the one to deal with. One answer only tells you about that one way of asking. Customers’ conditions cover the budget, who the present is for and what it’s for, and each time one of them changes, Gemini is working to a different set of conditions. So you have to check your prices, specs and stock against each way of asking, one at a time, to know whether you’re missing because you don’t fit the conditions or because you fit and weren’t picked.

Whether you’re in the recommendations for a question that a customer would ask is something that a demo with us will show you: one of our consultants will browse Gemini with you, come up with something that a customer would ask based on what you sell, like the exchange-gift request above, and put it to Gemini there and then. Which products it recommends, whether you’re in the comparison table and whether the price and reviews shown for your product are right will all be there on the screen. Of the information that AI can read, your website is a part that you control completely, so the consultant will also pull up your website’s GEO audit score; the audit is free on geoweb.tw and scores, across 12 dimensions, how well AI can understand your site.

If you want to carry on from there, the way we do it is called managed GEO: we revise the content on your website, we go through the places where other websites mention your products one by one and set them against your website, and every so often we put the same set of customer questions to ChatGPT, Gemini and other leading AI engines and keep each answer. Which ways of asking include you in the recommendations, which don’t, and how you’re described are all on record to compare.


Further reading