Your industrial buyer picked three suppliers before ever contacting you
The supplier shortlist now forms inside ChatGPT, weeks before the first email. Suppliers who aren't named there never reach the quote stage.
The part of the sale you no longer get to watch
For decades, the start of an industrial purchase was visible. A buyer called a rep, asked a colleague, flipped through a trade-show catalogue, emailed three suppliers they knew. You might lose the deal, but you knew the deal existed.
That stage now happens inside a chat box, and nobody tells you.
In a Semrush survey of 622 US B2B professionals, with manufacturing as the second-largest industry in the sample, 41% said they start vendor research inside an AI tool and only then turn to a search engine to validate what it told them. Among those who use AI, 92% said it shaped their vendor shortlist.
Read that second sentence again. This isn't about traffic. It's about who makes the list.
What actually changed
The shape of the buyer's question changed. It used to be a keyword query: "PET resin supplier Ohio". Now it's a full request, with context and criteria built in:
"I need a food-grade PET resin supplier, certified, able to deliver 20 tonnes a month to the Midwest. Who are the options and what are the trade-offs?"
The answer isn't ten blue links. It's three or four names, with reasoning attached. The buyer copies them into a spreadsheet and starts there.
The two suppliers who weren't named didn't lose the deal. They never learned it existed.
Why this hurts more in industry
Three traits of industrial buying make this shift especially expensive:
The cycle is long and the list closes early. Qualifying a raw material supplier takes months. The decision about who gets evaluated is made in the first week, the cheapest moment to influence and the hardest to reverse later.
The ticket is large and recurring. You don't lose an order. You lose a supply agreement that repeats every month, sometimes for years.
The market is fragmented and opaque from outside. In many raw material categories, buyers genuinely don't know every capable manufacturer. They depend on some source to assemble candidates, and that source stopped being the industry directory.
The most common misreading
Plenty of manufacturers look at this and conclude: "my product is too technical, AI can't recommend it".
That's half true, and the other half deserves honesty.
In very narrow specialties, the kind with five or six manufacturers worldwide, an experienced buyer already knows the names, and AI adds little. There, the impact is smaller.
But most industrial volume isn't like that. Commodities, semi-specialties, inputs with dozens of capable makers, standardised components, packaging, everyday chemicals: there the candidate list gets built from scratch, frequently, by people who don't know the market by heart. That's the ground where being cited decides.
And there's a second case that usually goes unnoticed: the new buyer. The analyst who took over the category last month and didn't inherit their predecessor's contact list. To them, every supplier is unknown, including the one that has fed the plant for ten years.
What determines who gets named
Assistants don't draw names from a hat. They assemble an answer from what they can read and corroborate about each company. Four things carry weight:
- Existing in readable form. Technical specification as text on the page, not locked inside a PDF or behind a registration form.
- Being confirmed by third parties. Industry association, distributor catalogue, technical publication, public registry, customer review. A company that only talks about itself is a company without evidence.
- Answering the question buyers actually ask, which involves application, volume, certification and lead time, not just a product name.
- Being findable in the right region. Industrial questions often carry geography, and the answer changes with it.
None of these is a trick. They're precisely the facts a competent buyer would demand. The difference is that they now have to sit in a form a machine can read, quote and defend.
Where to start this week
Before any project, run the cheapest diagnostic there is: open an assistant and ask the five questions your best customers asked before they became customers. Write down who appears.
If your company isn't among the names, you've just found out why that quote request you were expecting never arrived.
If it is, ask the follow-up: on what grounds? Because the answer may be citing you for the wrong product, the wrong application or the wrong region, and that costs business too.