On November 9, Google officially launched the Chinese-language version of Gemini’s shopping feature in Taiwan, aiming to solve this repetitive work. The approach in this update is simple: replace filters with conversation. Gemini automatically compiles responses that include both text and images, listing the price, reviews, stock, and purchase channels users care about most. The underlying technology is Google’s Shopping Graph. Google says this database contains more than 50 billion product listings and receives as many as 2 billion data updates per hour.
If you ask in a conversational way, what will you get?
Google’s officially released use cases in Taiwan are very specific. The company lists five scenarios in total, covering dehumidifiers, carry-on suitcases, gift exchanges, sneakers, and home outfit items, each with a clear budget and recommended quantity. For example, for buying appliances for rental living: “I want to buy a dehumidifier suitable for a Taipei studio apartment, with a budget of under NT$8,000, valuing energy efficiency and quietness; please recommend 3 to 5 models and compile a comparison table of specifications, prices, and reviews.”; the need to find a carry-on suitcase before a trip is written as “Next month I’m going on a five-day, four-night trip to Japan; please recommend 3 lightweight, easy-to-push 20-inch carry-on suitcases within a budget of NT$6,000, and compare size, weight, and price side by side.” There is a detail worth noting here: even the number of trip days is included, which means handing a usage condition like “five days and four nights” to the model for judgment.
Gemini’s responses come in two presentation formats. When users need visual references, they receive carousel cards with product images and purchase links; when they need to weigh several options, a specifications comparison table is included.
The Gemini app can provide price comparisons and purchasing information across the web, rather than only within Google’s own shopping pages. The company also notes that the shopping feature’s data source is the Shopping Graph, so the update frequency of product information corresponds to the update speed of this database.
A list of 50 billion entries, replacing 2 billion entries per hour.
The Shopping Graph supporting this feature is the foundation of the whole thing. In its May 2025 AI Mode shopping announcement, Google officially stated that this dataset covers more than 50 billion global products, sourced from international retail channels all the way to local shops. Each product includes ratings, price, color options, and availability, and more than 2 billion data records are refreshed on Google every hour.
The scope of this architecture isn’t limited to the Gemini app. In the same announcement, Google described that the shopping experience in AI Mode is also built on the Shopping Graph. After a user asks a question, it triggers multiple searches running in parallel: first clarifying “what conditions a backpack for long-distance travel in the rain should meet,” then recommending options with waterproofing and easy-to-access pockets based on those conditions. This process of “first understanding the requirement conditions, then going back to filter products” follows the same logic as the comparison tables in the Gemini app.
another link in the Google Shopping chain
If you look only at the comparison table in the Gemini app, you will underestimate the scope of Google’s shopping efforts over the past few years. Based on information officially released by Google, at least several related features have been rolled out one after another.
Virtual try-on is one of the earlier pieces. In a May 2025 announcement, Google said users can upload full-body photos of themselves to try on billions of apparel items in the Shopping Graph. The feature is powered by an image generation model built specifically for fashion, which understands details of the human body and clothing, such as how different materials fold, stretch, and drape on different body shapes.
Google’s push in shopping this year has had a clear main thread: turning its product database into a conversational interface. The underlying layer is a shopping graph of 50 billion listings updated with 2 billion entries per hour; Gemini and AI Mode are the entry points, while virtual try-on, agentic checkout, and a universal cart are extensions toward the transaction side. The piece Taiwan connected on November 9 brought the entry point to the entire chain in front of Traditional Chinese users.
For ordinary consumers, the most practical change is that it lowers the time cost of comparison shopping. In the past, you had to open five tabs and copy down six fields yourself to complete a comparison; now, with a single query—“20-inch carry-on suitcase within a budget of NT$6,000”—you can get a comparison table. As for when more advanced capabilities such as price tracking and automatic checkout will arrive in Taiwan, Google did not provide a timeline in this article.
Source: KOCPC Chinese