How to Use AI to Find Winning Products on 1688 (A 10-Minute Workflow)

An importer I know recently showed me his product research setup. He opens three browser tabs — ChatGPT, Claude, and DeepSeek — and runs through a sequence of prompts that turns a vague product idea into a shortlist of 1688 suppliers with estimated margins, in about ten minutes. He doesn't start product research any other way anymore.

I've been using a version of his workflow. The exact process goes like this.

The Workflow Overview

This is a four-step process that takes about 10 minutes per product idea:

1. Trend spotting (2 min) — Ask AI what's trending in your category
2. Demand validation (3 min) — Check if there's real demand on Amazon
3. Supplier matching (3 min) — Search 1688 with AI assistance
4. Margin estimation (2 min) — Calculate whether the numbers work

Step 1: Trend Spotting (2 minutes)

Start with a product category you're interested in. Instead of browsing 1688 randomly, ask an AI tool what's currently trending in that category on Amazon and TikTok Shop.

Prompt for ChatGPT or Claude:

I'm looking for trending products to import from China and sell on Amazon US.

Category: [kitchen tools / home organization / pet supplies / fitness accessories]

Give me:
- 3 specific subcategories with growing demand right now
- What customer complaints exist in each subcategory (based on Amazon reviews)
- The typical price range on Amazon
- Whether this subcategory is dominated by big brands or has room for new sellers

Be specific — name actual product types, not broad categories.

The output will give you a shortlist of 3-5 specific product ideas with context on competition and customer pain points. Pick the one that looks most promising.

Example result for kitchen tools:

The AI might tell you that "adjustable measuring spoons" have growing search volume, customer complaints focus on the sliding mechanism breaking after a few months, the average price on Amazon is $8.99-12.99, and the category isn't dominated by major brands. That's a specific product to pursue.

Step 2: Demand Validation (3 minutes)

Before you search 1688, check whether the product has enough demand to justify importing. You can do this with AI by analyzing Amazon listing data.

Prompt for Claude (better at analysis):

Analyze these Amazon listings for [product name] and tell me if this is a good category for a new importer:

- How many reviews do the top 10 listings have? (indicators of competition level)
- What's the average price? 
- What are common customer complaints?
- What features do top-rated products have that cheaper ones lack?
- Is there a specific feature gap you can identify?

Base this on common patterns in this type of product on Amazon US.

The goal here isn't exact data — it's pattern recognition. AI has seen enough Amazon listings across enough categories to identify common indicators. If the top 10 listings all have 5,000+ reviews, that's a highly competitive category. If most have under 500 reviews, there's room to enter.

Step 3: Supplier Matching on 1688 (3 minutes)

Now you know what product you want and that there's demand. The next step is finding suppliers on 1688.

Approach 1: AI-assisted 1688 search

Go to 1688.com and search for the product in Chinese. If you don't know the Chinese name, ask DeepSeek:

Translate this product name to Chinese for 1688 search:
[product name]

Also give me 3-5 alternative search terms in Chinese that 1688 suppliers might use.

DeepSeek is better than ChatGPT for this because it handles Chinese product terminology more accurately. For example, "adjustable measuring spoons" might be listed as "可调节量勺" on 1688, but suppliers might also use "伸缩量勺" or "多功能量勺". DeepSeek will catch these variations.

Approach 2: Supplier evaluation with AI

Once you have a list of suppliers, paste the product titles and prices into Claude:

I'm evaluating 1688 suppliers for [product name]. Here are my findings:

Supplier 1: [price, MOQ, location]
Supplier 2: [price, MOQ, location]
...

Which supplier should I order samples from first? Consider:
- Price vs. MOQ balance
- Location (which city — affects shipping to my consolidator)
- Payment methods accepted
- Any red flags in the listing descriptions

Step 4: Margin Estimation (2 minutes)

The final step is calculating whether the numbers work.

Prompt for any AI tool:

Calculate the estimated margin for this product:

Product: [name]
1688 unit price: $X
Estimated shipping per unit (sea freight): $X
Amazon FBA fee (est.): $X
Amazon selling price: $X
Estimated tariff (Section 301): 25% on product cost

Calculate:
- Total landed cost
- Amazon fees (15% referral + FBA)
- Net profit per unit
- Margin percentage
- Break-even monthly sales to make this worth doing

This won't be perfectly accurate — you'll need actual shipping quotes and FBA fee calculations for precision. But it's good enough to tell you whether a product is worth pursuing further.

Putting It All Together: A Complete Example

Here's how the full workflow plays out for one product idea:

Step	Action	Time	Tool
1	Ask for trending kitchen gadgets	2 min	Claude
2	Get "adjustable measuring spoon" as a candidate, analyze reviews	3 min	Claude
3	Search 1688 with Chinese terms from DeepSeek	3 min	DeepSeek + 1688
4	Calculate margin: $1.20 cost, $4.50 landed, $9.99 selling price, ~35% margin	2 min	ChatGPT

Total: 10 minutes to go from "I want to source kitchen tools" to "I have a specific product with a supplier shortlist and margin estimate."

Going Beyond the Basics

The workflow above is a starting point. Importers who use AI regularly develop their own prompt libraries and refine them over time. Common additions include:

• Competitive analysis prompts that analyze specific competitor products in detail
• Supplier communication templates in Chinese for initial inquiries (already covered in another article)
• Listing optimization prompts for Amazon and TikTok Shop product pages
• Customer service automation templates for common post-purchase scenarios

Go Deeper

The prompts used in this workflow are adapted from the AI Toolkit for Cross-Border Sellers, which includes 50+ prompts across 7 categories (product research, listing optimization, content creation, supplier communication, customer service, market analysis, and translation). It also includes 4 complete workflows — the one above is a condensed version of the product research workflow. If you use DeepSeek for supplier communication, the DeepSeek Prompt Pack has 200+ templates optimized for the same scenarios.