The shortlist looked impressive. Two of the five factories no longer existed.
A buyer recently showed us a supplier shortlist produced entirely by an AI tool. Well-formatted, confident, detailed. There was just one problem: when we checked, two of the five companies had deregistered, and a third had lost the certification the report praised.
The AI had not made anything up. It had simply read the internet — and the internet is a museum of how things used to be.
What AI research is genuinely good at
- Narrowing the field — from ten thousand possible suppliers to a dozen worth a closer look
- Mapping a category — which regions specialise in your product, typical price bands, common specifications
- Preparing better questions — knowing what to ask before the first conversation
What it cannot do
- Confirm the business licence is still active *today*
- Verify the certificate on the factory's website was issued to them — and not expired
- Walk the production line and see whether the machines match the brochure
- Tell you the "manufacturer" is actually a trading company renting a showroom
The right way to use it
Use research — AI or otherwise — to decide what to verify. Never treat the research itself as verification. The shortlist is the starting gun, not the finish line: every name on it still needs a licence check, a certification check, and eyes on the factory floor.
Want your shortlist verified by people who can actually visit? Send it to us — verification is what we do.*Warren Smith is the CEO of LS International. Our team in China verifies factories in person — because a spreadsheet can't.*