This article is for contract and custom manufacturers: machine shops, fabricators, molders and private-label producers that win work through requests for quotation (RFQs). The prize here is not website traffic. It is a place on a shortlist that can lead to years of production orders.
The short version
- Demand for outside production is strong: on Thomasnet, contract and private-label manufacturing was the top category buyers sourced in 2025 (opens in a new tab), across more than 1.5 million monthly sourcing sessions.
- Engineers use AI but check it: in the 2026 State of Marketing to Engineers (opens in a new tab) research by TREW Marketing and GlobalSpec, 69% of technical buyers used generative AI in purchasing, they rated its answers 4.7 out of 10 for trust, and 62% of the buying journey happened online before contacting a vendor.
- Supplier switching is creating openings: 32% of contract manufacturers were quoting reshoring projects in the 2026 Reshoring Initiative survey (opens in a new tab), double the 16% in 2025, and in Kearney’s survey (opens in a new tab) three in four respondents had moved sourcing from China to other low-cost countries.
- Value sits in a few large accounts: Xometry (opens in a new tab) had 81,821 active buyers at the end of 2025, and 1,760 accounts spending at least $50,000 a year.
- Brand recognition decides close calls: 70% of technical buyers said they were likely to choose the better-known brand when two solutions are technically similar.
Who buys from contract manufacturers, and what is a customer worth?
Engineers and sourcing managers at manufacturers, and one won account can mean years of repeat production orders.
The buyer is usually a design or manufacturing engineer who needs parts made, working with a sourcing or procurement manager who owns the supplier relationship. Thomas’ 2025 sourcing report found two-thirds of active buyers on its platform represent small and midsize businesses, with a significant share under age 45.
The supplier side is fragmented. According to the National Association of Manufacturers (opens in a new tab), there were 239,265 manufacturing firms in the US in 2022, and around three-quarters had fewer than 20 employees. Thomas says it connects buyers with more than 500,000 suppliers. Most of these firms have no marketing department, which is exactly why discovery matters: a buyer cannot shortlist a shop it has never heard of.
The money moving through this market is large. The US Census Bureau reported new orders for manufactured durable goods of $338.6 billion in August 2026 alone.
What one customer is worth varies by process and program, and there is no public benchmark for an average contract value. Public data shows the shape instead. At Xometry, the marketplace for custom manufacturing that also owns Thomas, active buyers grew 20% to 81,821 in 2025, while accounts spending at least $50,000 over twelve months grew 18% to 1,760. Our inference: a small share of accounts carries much of the value, so one well-qualified RFQ can matter more than thousands of website visits.
Where does AI already sit in supplier discovery?
In early research and longlisting, used by a majority of technical buyers but trusted only after checking.
The best industry-specific evidence is the 2026 State of Marketing to Engineers survey from TREW Marketing and GlobalSpec, with Elektor:
| Technical buyers, 2026 (TREW Marketing and GlobalSpec) | Share |
|---|---|
| Use generative AI during the purchasing process | 69% |
| Never use it to evaluate vendors or research options (42% in 2025) | 31% |
| Use it often (6% in 2025) | 11% |
| Notice AI summaries at the top of search results | 75% |
| Of those, read the summary and then continue to the results | 69% |
| Of those, find the summary usually enough | 6% |
| Routinely research in online technical publications | 76% |
| Routinely research on supplier and vendor websites | 74% |
Two findings matter most for a supplier. Average trust in generative AI answers was 4.7 out of 10, so engineers use AI to start a search and then verify on supplier websites and in trade publications. And the online share of the journey keeps rising: buyers 35 and under complete 66% of the process online before their first vendor conversation. The same habit shapes how engineers choose components with AI.
Industrial platforms are moving the same way. In January 2026, Thomas launched an AI search that lets buyers run detailed, multi-attribute searches in natural language. Digital Commerce 360 reported (opens in a new tab) that in testing it drove more than 15% more supplier evaluations than the old search.
Procurement teams are earlier in adoption. In EFESO’s survey of European procurement, 79% of respondents had been exposed to AI, but only 31% actively used generative AI tools for work. Across business buying generally, Forrester (opens in a new tab) describes generative AI searches as the starting point for B2B buyers, followed by validation from trusted people and sources.
Which supplier questions do buyers ask AI?
Questions that combine a process, a material, a standard, a volume and a place, the same filters an RFQ uses.
The prompts below are illustrative, written by us to show the shape of supplier questions. They are not captured from real buyers.
| Need | Illustrative prompt |
|---|---|
| Capability and certification | “AS9100 certified machine shops for 5-axis titanium parts, low volume” |
| Regulated production | “ISO 13485 injection molders for medical device housings in the Midwest” |
| Reshoring | “US alternative to an overseas supplier for stamped steel brackets, about 50,000 a year” |
| Private label | “Contract manufacturers for private-label skin care with low minimum orders” |
| Speed | “Sheet metal fabricators that can turn prototypes in under two weeks” |
| Comparison | “Compare these three molders on capacity, certifications and industries served” |
Each prompt is a filter. If your site, directory profiles and trade coverage never say that you hold a given certification, run a given process or serve a given region, an assistant has nothing to match. The Thomas data shows which filters are active: buyers in aerospace and defense sought precision CNC machining and fabrication, food and beverage buyers sought contract manufacturing for supplements and vitamins, and the sharpest growth came in military equipment, automation, robots and prototyping services.
How does an AI answer become an RFQ and a contract?
Through a longlist, a website check, a quote competition and qualification, then repeat production orders.
- Longlisted. An engineer asks an assistant or an industrial platform for suppliers that match the job. The AI answer is a starting list, not a decision.
- Checked. The engineer visits supplier websites and trade publications to confirm capabilities, certifications, equipment and industries served. TREW and GlobalSpec found buyers reach out when they need what they cannot find online: pricing and availability.
- Asked to quote. The supplier receives an RFQ, usually alongside rivals. In the Reshoring Initiative’s survey, reported by Floor Covering News (opens in a new tab), contract manufacturers competed against imports on an average of 38% of quotes, and when they lost to imports, price was the main factor 94% of the time.
- Qualified and awarded. Samples, first-article inspection and audits follow before production starts.
- Repeated. A qualified supplier tends to keep receiving orders for the life of the program. That is where the value of the first mention is realized.
AI visibility affects the first two steps. Price, quality and delivery still decide the last three. A shop that is never longlisted never gets the chance to compete on them. Equipment makers chasing a place on engineers’ bid lists face the same first cut.
What decides whether an assistant names your plant?
Facts it can find and confirm in outside sources; the platforms describe how they search, not how they choose.
What is documented by the platforms: OpenAI says (opens in a new tab) ChatGPT search rewrites a question into one or more targeted queries sent to search providers, and that a site must allow OpenAI’s search crawler, OAI-SearchBot, to be eligible for inclusion. Google says (opens in a new tab) AI Mode uses a “query fan-out” technique, running multiple related searches across subtopics. Neither publishes how suppliers are chosen.
What has been observed:
- Assistants run several searches per question. In our study of hidden searches, ChatGPT ran a mean of 3.7 searches per buyer question before answering.
- Independent coverage matters most of what we measured. In our brand entity study, each tenfold increase in independent sites naming a brand in the cited pages went with 4.7 times the odds of being recommended.
- Answers vary. In our consistency study, only 25.2% of the brands ChatGPT named for a question appeared in all five repeats.
- Familiarity tips decisions. In the engineers survey, 53% said brand familiarity influenced their most recent purchase.
Our inference for contract manufacturing: the trust factors are the ones a sourcing manager already checks, stated where machines can read them. That means certifications with the registrar named, process and equipment lists, materials, tolerances, volume ranges, industries served, plant locations, export-control registrations where they apply, and coverage in trade publications and industry directories. Chemical producers face a similar checklist, set out in how chemical suppliers reach B2B buyers.
What does a contract manufacturer lose when AI leaves it out?
It loses RFQs it never sees, at a time when many buyers are actively looking for new suppliers.
We have no direct measurement of RFQs lost to AI absence, so we label the reasoning:
- Buyers are moving work. In the Reshoring Initiative survey, 36% of original equipment manufacturers (OEMs) had reshored or were actively reshoring in 2026, and 79% of contract manufacturers said at least some customers had discussed reshoring with them. Kearney found most companies moved sourcing to other low-cost countries rather than home: only 20% of its survey respondents had looked at domestic manufacturing. Either way, supplier searches are happening.
- The search happens before contact. With 62% of the journey online, a supplier missing from the longlist is not in the comparison at all. We infer the loss shows up as RFQs that never arrive, which no CRM records.
- Wrong facts are costly. An assistant that says you lack a certification you hold, or places you in the wrong state, filters you out. Correcting a missing certification or a wrong plant location starts with the page the assistant read; our guide to fixing wrong brand information in AI answers walks through it.
How does GEO work for a contract manufacturer?
Generative engine optimization (GEO) makes your plant easy for AI assistants to find, describe accurately and verify.
For a contract manufacturer, the work usually covers:
- Capability pages in plain text. One page per process with materials, tolerances, part sizes, volumes, secondary operations and lead-time ranges, written as text, not only in brochures, PDFs or photos.
- Certifications you can prove. Each certification named with its scope, registrar and expiry, matching the registrar’s own listing.
- One consistent identity. The same company name, plant addresses, capabilities and certifications on your site, Thomasnet, LinkedIn, association directories and customer supplier portals.
- Independent coverage. Trade publication features, association memberships, case studies published with customer permission, conference talks and supplier awards. For why outside sources matter, see how small brands get recommended by AI.
- Comparison-ready content. Clear answers to how you compare with offshore sourcing on lead time and total cost, without naming rivals unfairly. Our review of whether comparison pages help B2B brands get cited covers the limits.
- Crawl access. Make sure your site allows the search crawlers the assistants document, and that key pages load without logins.
- Measurement. Ask a fixed set of capability, certification and location questions across ChatGPT, Gemini, Perplexity, Copilot and Google’s AI features, repeatedly, and track who is named and which sources are cited. Then match those findings to RFQs received.
None of this guarantees a mention, and no honest provider can promise one. It makes your plant the easiest one for an assistant, and then an engineer, to verify. Hardware companies face the same verification test with their own buyers; makers of jobsite drones, robots and sensors are one example, covered in how construction technology reaches contractor shortlists.
Which questions can’t the data answer for manufacturers yet?
It shows engineers using AI in research, not how many RFQs or contracts AI answers create.
- Usage is not attribution. The TREW and GlobalSpec survey measures how buyers research, not which supplier choices came from AI. We found no public data linking AI answers to RFQ volume.
- Surveys come from interested parties. TREW sells marketing to engineering firms, GlobalSpec sells advertising, Thomas and Xometry run marketplaces, and the Reshoring Initiative promotes reshoring.
- No manufacturing-specific ranking studies. Our studies covered buyer questions across several industries, not supplier searches. Applying them to contract manufacturing is our inference.
- Deal values are private. Public figures show account counts and spending thresholds, not typical contract sizes for a job shop.
Where should a contract manufacturer start?
Start by asking assistants the supplier questions your best customers would ask, and see whether you appear.
That first check usually shows whether your plant is named for your core processes and certifications, whether your locations and capabilities are described correctly, which directories and publications the answers rely on, and which rival shops are listed instead. The work then is to state your capabilities where assistants read them and earn the outside coverage that confirms them.
If your growth depends on a handful of new production accounts a year, ask us to review your supplier visibility. We will show where your plant appears when buyers ask AI for suppliers, why other shops are named instead, and which changes are most likely to bring more qualified RFQs. Those changes, such as capability pages per process, provable certifications and one consistent plant identity, are what our generative engine optimization service page covers.
Frequently asked questions
Do engineers really use ChatGPT to find suppliers?
Many use generative AI somewhere in the purchase: 69% in the 2026 TREW and GlobalSpec survey. Most still verify on supplier websites and in trade publications before contacting anyone.
Does being on Thomasnet still matter?
Yes. Thomas reports more than 1.5 million monthly sourcing sessions and now offers natural-language AI search. Directory profiles are also outside sources an assistant can find, so keep them complete and consistent.
Can a small job shop compete with large contract manufacturers in AI answers?
It can be named for specific combinations of process, material, certification and location, where fewer suppliers match. Large firms still benefit from recognition, and 70% of technical buyers lean toward the better-known brand.
Will AI search replace our sales team?
No. Buyers contact suppliers for pricing, availability and complex requirements. AI changes who gets contacted, not the need for quoting and engineering conversations.
Sources
- Distribution Strategy Group (2026-01-30), Industrial Sourcing Behavior Shifts in 2025 (Thomas 2025 Annual Sourcing Activity Report) (opens in a new tab)
- TREW Marketing and GlobalSpec (2026), State of Marketing to Engineers research report (opens in a new tab)
- Digital Commerce 360 (2026-01-19), Thomas rolls out AI search and performance-based ads to speed industrial sourcing (opens in a new tab)
- Xometry via GlobeNewswire (2026-02-24), Xometry Reports Record Fourth Quarter and Strong Full Year 2025 Results (opens in a new tab)
- National Association of Manufacturers (2026), Facts About Manufacturing (opens in a new tab)
- US Census Bureau (2026-09-25), Advance Report on Durable Goods Manufacturers’ Shipments, Inventories and Orders, August 2026
- Design News (2026), US Reshoring Accelerates Despite Policy Uncertainties and Skilled Worker Shortages (opens in a new tab)
- Floor Covering News (2026-09), Reshoring survey finds US manufacturing momentum building (opens in a new tab)
- Kearney (2026), 2026 Reshoring Index (opens in a new tab)
- EFESO (2026), The State of Generative AI in European Procurement
- Forrester (2026-01-21), Forrester’s 2026 Buyer Insights: GenAI Is Upending B2B Buying (opens in a new tab)
- OpenAI Help Center (2026), ChatGPT search (opens in a new tab)
- Google (2025-03-05), Expanding AI Overviews and introducing AI Mode (opens in a new tab)
- Underneath (2026), The hidden searches AI assistants run before they answer
- Underneath (2026), Do Wikipedia and schema make AI assistants recommend a brand?
- Underneath (2026), Ask an AI the same question 5 times: do the brands change?