For AI

Answers

How do I export Taobao and Tmall product reviews for analysis?

Short answer

Find product IDs with okdata's Taobao product search, about 10 products a page for $0.043, then pass each ID to Taobao product reviews: about 20 reviews a page for $0.043, newest first, with stars, text and date. Ten pages (about 1,000 reviews) on each of five competitors cost $2.193.

Taobao (淘宝) and Tmall (天猫) are Alibaba's two big marketplaces in China: Taobao for every kind of seller, Tmall for brand and flagship stores. They share one search and one product ID system. If you are launching a product in China or watching a competitor, the most useful material is often in other sellers' reviews: what buyers complain about, what they praise, what they compare it with. Reviews can only be read by scrolling a product page, and there is no export. The approach: find the products by keyword, then pull their reviews by product ID into a sheet for a person or a language model to read.

The calls

  1. Find the products. Taobao product search takes Chinese keywords in query and sorts by sales by default; tmall_only=true keeps Tmall stores only, and price_min and price_max set a price range in yuan. Each result has the product id, title, price, sold (a band for buyers in the last month) and the shop name. If you already have a link, the number after id= in an item.taobao.com or detail.tmall.com URL is the product ID and you can skip this step.
  2. Pull the reviews. Pass the ID to Taobao product reviews. Each review has a rating of 1 to 5 stars, text and published_at. sort defaults to recent, newest first, or default for Taobao's own order. Page with page.next_cursor.
  3. Variants and prices, when needed. Taobao product details returns the images, item properties and every SKU with its price and stock, $0.033 a product.

A real run

On November 20, 2026 we searched Tmall for 挂耳咖啡 (drip-bag coffee) and got one page of 10 products. sold took only two values, 10000 and 1000, which are the "10,000+" and "1,000+" bands, and rating and reviews were empty. One of them:

{
  "id": "706433184157",
  "title": "【同价双11】瑞幸即享挂耳咖啡深度烘焙美式黑咖啡72袋囤货装",
  "price": 259,
  "sold": 10000,
  "shop": { "name": "瑞幸即享咖啡旗舰店" }
}

That is a 72-bag pack from Luckin Coffee's Tmall store at 259 yuan. Its first page of reviews held 20, all from November 18 and 19: 18 five-star and 2 one-star. Eleven of them read 该用户觉得商品非常好,给出5星好评 ("this user found the product very good and gave five stars"), the default text Taobao fills in when a buyer leaves no words. One written review:

{ "rating": 5, "text": "挂耳袋不堵塞。 烘焙度有点偏深。 分量挺足的。", "published_at": "2026-11-19T09:03:00.000Z" }

("The filter bag doesn't clog. The roast is a bit dark. Generous amount.") variant (the option bought), images and likes were not returned in this run.

What it costs

Approach Calls Cost
One search page to find products 1 $0.043
Ten pages of reviews on one product (about 200) 10 $0.43
Five competitors, ten pages each, plus one search 51 $2.193

Gotchas

  • Remove default reviews first. The auto-filled five-star text says nothing, and here it was 11 of 20. Filter it out before sentiment analysis or the share of positive reviews comes out too high.
  • Sales are bands. sold is a "1,000+" or "10,000+" figure, good for rough comparisons only.
  • Buyer names are masked. Taobao shows only part of the name and author_name is the same; analysis does not need it.
  • Public reviews only. Anything on the product page can be pulled. A shop's refund reasons or customer service chats stay in the seller's own back office.
  • New products have few reviews. Stop when has_more is false.

Sources

Where people asked this (read 2026-11-20)

Updated

Connect once, then just ask.

Every new key comes with $0.10 to try it. One sentence connects your agent.

Start free
Connect your AI

Paste this into your agent

Read https://okdata.online/agents.md and set up okdata for me; my key is in the OKDATA_API_KEY environment variable.

Claude Code, Cursor, Codex and other agents that can open a link set themselves up from it.

MCP serverhttps://api.okdata.online/mcp