Blog

How do overseas buyers use ChatGPT to find suppliers in China?

By Dawn, Founder of CitedAsia · Updated 28 October 2026

Shipping containers and cranes at a busy container port
Short answer: Overseas buyers use ChatGPT and similar tools to build a first shortlist of suppliers, compare options and check a supplier before contacting them. They type detailed prompts with product, material, certification, MOQ and region, and the AI answers from web pages it can find and read. Exporters who publish specific, verifiable facts in English on pages AI tools can access are far easier to shortlist than those with only a brief platform listing.

Do overseas buyers really use ChatGPT to find suppliers?

A growing share do, but it varies by industry and market, and search engines and B2B platforms are still part of the journey. The best recent data point is a 2025 study by Responsive, a software company, which surveyed 350 B2B buyers. It found that about a quarter said generative AI has overtaken traditional search for them, and nearly two-thirds use it as much as or more than search when researching vendors. The same study, as reported by Digital Commerce 360, found big differences: 48% of US buyers use generative AI to find vendors, against 14% in other regions, and adoption was much higher among technology buyers.

Treat this as a direction of travel, not a precise measure for manufacturing. The sample is cross-industry and the publisher sells software. Two more data points show the scale:

What do buyers type into AI tools when sourcing?

Buyers write long, specific prompts, much closer to an RFQ than to a keyword search. The examples below show the typical stages. They are illustrative, not quotes from real buyers.

StageExample promptWhat the AI needs to find
Discovery“Manufacturers in China of food-grade silicone bakeware with LFGB test reports, MOQ under 1,000 pieces”Product pages naming the material, the standard and the MOQ
Shortlist“Compare three suppliers of stainless steel vacuum flasks for private label, with lead times”Lead times, OEM/ODM options, capacity
Verification“Is [company] a factory or a trading company? Where is it located?”About page with factory address, ownership, year founded
Compliance“Which certifications does [company] have for selling into the EU?”A certification page with standards, certificate numbers and issuing bodies
Logistics“Suppliers near Ningbo port who can ship FOB within 30 days”Location, port, Incoterms you offer, typical production time

Look at the right-hand column. Nearly every answer depends on facts that many exporters only share privately, after an enquiry. If those facts aren’t published, the AI can’t use them.

Workers on a manufacturing production line inside a factory
Buyers want to know who actually makes the product, and where. (Photo: Unsplash)

Which sources do AI tools cite when recommending suppliers?

They cite pages they are allowed to access and can understand, from both your own site and third parties. Each tool has its own crawler. OpenAI says OAI-SearchBot is used to surface websites in ChatGPT search, and sites that opt out of it will not be shown in ChatGPT search answers. GPTBot is a separate crawler for model training. Perplexity says PerplexityBot is used to surface and link websites in its results. Google says AI Overviews and AI Mode use its normal search systems, and a page must be indexed and eligible to show with a snippet.

In practice, sourcing answers tend to draw on a mix of:

  • Suppliers’ own websites, especially product and About pages written in clear English
  • B2B marketplace listings
  • Trade show exhibitor lists, industry association directories and trade media
  • Comparison or “top suppliers” articles

You can’t control third-party articles, but you can make sure the facts about your company are the same everywhere and that your own site is the most complete source.

How an overseas buyer moves from an AI prompt to an RFQ A four-step flow: the buyer asks AI for suppliers, the AI builds a shortlist from pages it can read, the buyer verifies the shortlisted suppliers, then sends an enquiry. Under each step is what the exporter needs to have in place. From prompt to RFQ: where your website does the work 1. Buyer asks “Factories in China for LFGB-tested silicone bakeware, low MOQ” 2. AI shortlists Searches the web, reads pages it can access, cites a few suppliers 3. Buyer verifies “Is this a real factory? Which certificates? Lead time?” 4. RFQ sent Email or form, to 2-5 suppliers (example) What you need at each step • Product words buyersactually use • Material, standard,MOQ in plain text • English pages foreach product line • Search crawlers allowedin robots.txt • Specs in HTML, notonly in PDFs or images • Mentions on trade showand association sites • Factory address, photos,year founded • Certificate numbersand issuing bodies • Same company nameon every platform • Clear RFQ form • Business email onyour own domain • Stated reply time • English-speakingcontact person Example product and prompt for illustration
A simplified buyer journey. The AI only shortlists suppliers it can find and read, and the buyer only sends an RFQ to suppliers who pass verification.

What does this mean for Chinese exporters?

Your website must answer the buyer’s prompt before the buyer contacts you. Five practical steps:

  1. Be specific. Name the materials, standards, MOQ, capacity and lead times on each product page. The GEO study by Aggarwal et al. (KDD 2024) found that adding statistics, quotations and citations improved visibility in generative engine answers by up to about 40% in the authors’ tests.
  2. Be verifiable. List certificate numbers and issuing bodies, your factory address and when you were founded. If you are a trading company, say so plainly and explain which factories you work with. Buyers ask, and misleading answers cost trust.
  3. Be accessible. Check that your robots.txt does not block OAI-SearchBot, PerplexityBot or Googlebot, and that key specs are in HTML text, not only in PDF catalogues or images.
  4. Be consistent. Use the same English company name, address and product names on your website, B2B platform stores, LinkedIn and trade show listings.
  5. Be mentioned. Exhibitor pages, association memberships and trade media coverage give AI tools independent confirmation that your company is real.
Try this today: Write three prompts a real buyer of your main product would type, in English, with material, certification and MOQ. Run them in ChatGPT (with search on) and Perplexity. Write down which suppliers are named and which pages are cited. Then open your own product page and check whether it states the same facts as the pages that were cited.

Frequently asked questions

Should I block GPTBot on my export website?

That is a separate choice from search visibility. OpenAI says GPTBot is used for model training, while OAI-SearchBot is used to show sites in ChatGPT search. You can block one and allow the other. If you want to appear in ChatGPT search answers, don’t block OAI-SearchBot.

Do AI tools prefer factories over trading companies?

There is no published rule. Buyers often ask whether a supplier is a factory, so state clearly what you are. An honest answer is more useful than a vague one.

Is a PDF catalogue enough for AI tools?

Don’t rely on it. Put your key specs, MOQ and certifications on HTML product pages too, so search crawlers and buyers can read them easily.

Do I need content in languages other than English?

Start with English for international buyers. Then add the languages of your main target markets, so buyers who search in their own language can find you.

Want to see how ChatGPT and Perplexity answer your buyers’ sourcing questions today? Request a free AI visibility snapshot, or read more about GEO for Chinese exporters (中文版).

Sources: Responsive, 2025; Digital Commerce 360; TechCrunch; MIT Technology Review; OpenAI crawlers; Perplexity crawlers; Google Search Central; Aggarwal et al., KDD 2024.

Is your business in the answer?

Free check of 5 customer questions on ChatGPT and Perplexity. Score in 30 seconds, full report within 24 hours (business days).