• AI visibility

Guide · AI search

Can AI search bring our CNC shop the RFQs that now go to instant-quote platforms?

It can, but only for jobs where an assistant can confirm that your shop holds the tolerance, runs the material, carries the certification and sits in the right place. The instant-quote platforms already publish those facts in a form machines can read; most job shops still keep them in a brochure or in the owner’s head.

By Underneath Team Guide Updated

This guide is for owners and sales leads of precision CNC machining shops: milling, turning, 5-axis and Swiss work for aerospace, defense, medical, semiconductor and industrial customers. Our broader article on how contract manufacturers win RFQs through AI search covers reshoring and general supplier discovery. This one stays with the machine shop and its specific rival: the online platform that quotes a part in minutes.

The short version

  1. The platforms are growing fast: in the second quarter of 2026, Xometry (opens in a new tab) grew marketplace revenue 45% to $215 million and active buyers 20% to 89,557, and Protolabs (opens in a new tab) grew CNC machining revenue 13.6%.
  2. Machine shops are investing, but lagging: the AMT’s order report (opens in a new tab) put US metalworking machinery orders at $2.77 billion in the first five months of 2026, up 31.9%, while orders from contract machine shops ran nearly 10% below their recent average in May.
  3. Buyers find sourcing painful: in the 2026 Fictiv and MISUMI survey (opens in a new tab) of more than 300 manufacturing leaders, 81% said supplier sourcing is too time-consuming and costly, up from 73% in 2025.
  4. Engineers are starting to use AI, cautiously: in the 2026 State of Marketing to Engineers (opens in a new tab) survey, 21% routinely used generative AI platforms in purchase research, up 8 points, and they rated their trust in AI answers 4.7 out of 10.
  5. Defense work now has a new gate: the CMMC cybersecurity rule took effect on November 10, 2025, and from November 10, 2026 applicable defense contracts can require a third-party Level 2 assessment.

Who buys CNC machining, and what is one account worth?

Engineers and sourcing managers at manufacturers, and a good account is worth years of repeat part orders.

The person who starts the search is usually a mechanical or manufacturing engineer with a drawing and a deadline. A buyer or sourcing manager then owns the supplier relationship, the quote comparison and the purchase order. In aerospace, defense and medical work, a quality engineer also checks the shop’s certifications before the first article ships. Chemical buyers run a similar document check, described in how formulators find chemical suppliers.

The industry is investing. According to the US Manufacturing Technology Orders report from AMT, the Association For Manufacturing Technology, metalworking machinery orders reached $583.4 million in May 2026, 47.8% more than a year earlier. AMT calls contract machine shops “the largest customer of manufacturing technology by industry,” but notes their orders have been falling behind total orders since December 2025, with much of the growth coming from aerospace. The builders selling those machines face their own AI shortlist, covered in how machinery makers get shortlisted.

There is no public benchmark for what an average job shop account is worth. The platforms show the shape of the value instead. At Xometry, accounts spending at least $50,000 over twelve months grew 23%, from 1,653 to 2,039, in the year to June 2026. Our inference: a shop’s revenue rests on a few dozen repeat accounts, so one well-matched RFQ from a new program can matter more than a month of website traffic.

Where do instant-quote platforms and AI search meet the buyer?

At the start of a job, when the engineer wants a fast price or a short list of capable shops.

Instant-quote platforms answer the engineer’s first question, price and lead time, without a phone call. Xometry describes itself as an “AI-native marketplace” and, in its second-quarter 2026 report, said it launched a process recommender that suggests the best of 20 manufacturing techniques when a buyer uploads a part. It also scores each job against a supplier’s “machine characteristics, quality history and on-time shipping record.” In other words, the platform is doing the matching a buyer used to do by phone.

Buyers like that convenience. In the Fictiv and MISUMI survey, 97% of leaders called digital manufacturing platforms essential for production, up from 86% in 2024. Fictiv is itself a manufacturing platform, so read that figure as a signal from an interested party, not a neutral count.

General AI assistants are a second, newer route, and the engineers who send RFQs are only part way there. In the TREW Marketing and GlobalSpec survey of more than 1,000 engineers and technical buyers, 69% use generative AI somewhere in the purchasing process, but the share who make it a routine part of purchase research is 21%, up 8 points. About 31% routinely use industry directory websites, and 77% still use a search engine more often than an AI platform. Engineers do 62% of the buying process online before contacting a vendor, and for 26% the first contact with sales is driven by pricing or inventory questions.

Two points stand out for a job shop. Search engines still dominate, so the pages that rank for capability searches are also what many assistants read. And the first reason engineers contact a vendor is price and availability, which is exactly what instant-quote platforms supply without contact.

Directories are adding AI too. In January 2026, Thomas launched an AI search that lets buyers run 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.

What do engineers ask when they look for a machine shop?

Questions that stack a process, a material, a tolerance, a certification, a quantity and a place.

The sample questions below are ours, built to show how a drawing turns into a search. They are not captured from real buyers or from AI logs.

What the buyer needsExample question (illustrative)
Aerospace capability“AS9100 machine shops with 5-axis capacity for aluminum housings near Wichita”
Defense eligibility“ITAR-registered CNC shops that can machine titanium brackets, 50 pieces”
Medical precision“Swiss turning shops holding half a thousandth on 316 stainless pins, ISO 13485”
Cybersecurity status“CMMC Level 2 machine shops in Ohio for defense parts”
Platform or local shop“Should I use an online quoting service or a local machine shop for 200 production parts?”
Speed“Machine shops that can deliver prototype parts in five days”

Each filter removes most of the market. Our own research suggests the details matter to the answer, not just to the buyer. In our prompt phrasing study, adding “on a tight budget” to a question kept the original first brand only 15.3% of the time, against 68.0% when the same question was simply asked again. That study covered consumer and business questions, not machining, so applying it here is our inference: a tolerance, material or certification in the question will likely reshuffle which shops are named.

Location matters in a documented way. OpenAI says (opens in a new tab) ChatGPT search rewrites a question into targeted searches and may use the general location from a user’s IP address to localize them. In our study of four assistants, recommendations for questions that named a place overlapped between assistants far less than national ones, 0.160 against 0.390 on a scale where 1 is identical. Local shop answers are less settled, which means a well-documented regional shop has room to appear.

How does a shop named by an assistant end up with the purchase order?

Through a shortlist, a capability check, an RFQ, a first article and then repeat orders.

  1. Named or not. An engineer asks an assistant, a directory or a platform who can make the part. A shop that is not named is not on the list.
  2. Checked. The engineer opens the shop’s site to confirm machines, materials, tolerances and certifications. Thin pages end the visit here.
  3. Asked to quote. The RFQ goes to two or three shops, often alongside an instant quote from a platform as a price reference.
  4. First article. The winning shop ships first-article parts with inspection reports. Regulated programs add a quality audit.
  5. Repeat orders. A shop that passes keeps the part number for revisions and production. This is where the value of the first mention shows up.

Showing up in an assistant’s answer gets a shop through the shortlist and the capability check, and no further. Price, lead time, quality and responsiveness still win the job. What visibility changes is the mix of RFQs: a shop described clearly for its hardest work is more likely to be asked about that work, not just the commodity parts where platforms win on speed. For the wider supplier search, including reshoring, read how contract manufacturers win RFQs.

What makes an assistant name one machine shop over another?

Facts it can find and check outside your own claims; the platforms document how they search, not how they pick.

Documented by the platforms: ChatGPT search sends rewritten queries to search providers, and a site must allow OpenAI’s search crawler, OAI-SearchBot, to be eligible for inclusion. Google explains (opens in a new tab) that AI Mode splits a question into subtopics and runs many related searches, a method it calls “query fan-out,” so a single machining question can become separate searches for material, tolerance and location. Neither company publishes how it chooses a supplier.

Observed in studies:

  • Facts drift where sources disagree. In our business facts study, 18.9% of AI answers about local businesses stated at least one address, phone, website or hours fact that differed from the Google profile. A shop with an old address on one directory and a new one on its site invites that mistake.
  • Engineers verify. With trust in AI answers at 4.7 out of 10, the TREW and GlobalSpec data suggest an assistant’s mention starts a check, not a purchase.
  • Familiarity breaks ties. In the same survey, 70% said they were likely to choose the better-known brand when two solutions look technically similar.

Our inference for machine shops: the trust signals are the ones a quality engineer already asks for, published where software can read them. That means certificates with scope and registrar, a real equipment list, materials and tolerance ranges, part size limits, quantities, inspection equipment, and the industries you serve. For defense work, it also means stating your export-control and cybersecurity status accurately. Under the International Traffic in Arms Regulations (opens in a new tab), anyone in the US manufacturing defense articles must register with the State Department even if it never exports, and the first-tier fee is $3,000 a year. Under the CMMC rule, explained by Squire Patton Boggs (opens in a new tab), primes must flow requirements down to subcontractors that handle controlled defense information, and status is recorded in the government’s Supplier Performance Risk System. Claiming a status you do not hold is a compliance problem, not a marketing one.

What does a machine shop lose when assistants leave it out?

It loses the RFQs that never arrive, and it cedes the easy first quote to the platforms.

We have not found any measurement of RFQs lost to AI absence, so the reasoning below is labeled:

  • The platforms are taking share. Xometry’s chief executive said the company aims “to rapidly penetrate the vast, fragmented offline market.” Its buyer count and its large accounts both grew faster than the machine shop order trend AMT reports. We infer that part of that growth comes from work that once went to independent shops by phone and referral.
  • Buyers want less effort. With 81% of manufacturing leaders saying sourcing takes too much time and money, the supplier that is easiest to verify has an advantage before quoting starts. Packaging suppliers chasing quote requests face the same test.
  • Absence is silent. An engineer who builds a shortlist from an assistant and three websites never calls the shop that was missing. No CRM records the loss. To see how fewer search clicks show up in pipeline, read what lost clicks to AI answers mean for revenue.

There is also a dependence question. Shops that take overflow work as platform suppliers gain volume but not the customer relationship. Our article on whether AI search reduces dependence on marketplaces looks at that trade-off in another industry; we infer the same tension applies to machining.

How does GEO work for a CNC machine shop?

Generative engine optimization (GEO) makes your shop easy for AI assistants to find, describe correctly and verify.

For a machine shop, the work usually covers:

  1. Capability pages in plain text. A page for each process (milling, turning, 5-axis, Swiss, EDM, grinding) listing machines, axis counts, travel limits, materials, typical tolerances, quantities, secondary operations and lead-time ranges.
  2. Certifications with proof. ISO 9001, AS9100, ISO 13485, ITAR registration and CMMC status stated exactly, with scope and dates, and matching what registrars and customers can look up.
  3. One consistent identity. The same company name, address, phone and capability list on your site, Thomasnet, LinkedIn, Google Business Profile, association directories and customer supplier portals. Our guide to fixing wrong brand information in AI answers shows how errors spread.
  4. Independent coverage. Features in trade publications, a profile in association directories, published case studies with customer permission, and talks at shows such as IMTS. Engineers rank online technical publications as their top research source, and assistants read the same pages. A 30-person shop can still be named; our piece on how small brands get recommended by AI explains what makes that happen.
  5. An honest answer to the platform question. A page that explains when a local shop beats an online quote, for example on tight tolerances, complex setups, inspection documentation or ongoing engineering support, and when it does not.
  6. Crawl access. Allow the search crawlers the assistants document, and keep capability pages outside logins and PDF-only downloads.
  7. Measurement. Ask a fixed set of process, material, tolerance, certification and location questions across ChatGPT, Gemini, Perplexity, Copilot and Google’s AI features, repeatedly, and log who is named and which pages are cited. Then compare with RFQs received.

No one can promise that an assistant will name your shop. The aim is to make your shop the easiest one to confirm when it is a real fit.

What can’t the data tell a machine shop owner yet?

How many RFQs AI answers create, and how assistants treat machining questions specifically.

  • Usage is not attribution. The engineer survey measures research habits. We found no public data linking AI mentions to RFQs or purchase orders.
  • Interested sources. TREW sells marketing to technical firms, GlobalSpec sells advertising, Fictiv and Xometry run manufacturing platforms, and Protolabs competes with job shops. Their numbers are useful but not neutral.
  • No machining-specific answer studies. Our studies tested buyer questions across several industries, not machine shop searches. Applying their findings to CNC work is our inference.
  • Account values are private. Platform reports give thresholds such as $50,000 a year, not typical job shop contract sizes.

Where should a CNC machine shop start?

Ask assistants for shops that can do your best work, and check whether you are named and described correctly.

Pick the ten jobs you most want more of, phrase them the way an engineer would, with material, tolerance, certification and region, and run them across the main assistants. Note whether you appear, what they say about your capabilities, and which shops and platforms show up instead. That usually shows which facts are missing from your site and which outside sources the answers lean on.

If you would rather quote fewer commodity parts and more of the tight-tolerance work your machines were bought for, ask us to check how assistants describe your shop. We run the process, material and certification questions your buyers ask, show which shops and platforms are named in your place, and list the capability facts that are missing or wrong. To see how fixing those facts is handled, from process pages to certification proof and directory cleanup, read about our generative engine optimization service.

Frequently asked questions

Will instant-quote platforms replace local machine shops?

The data does not show that. Platforms are growing fast, and Xometry also routes work to partner shops. Complex, tight-tolerance and documented regulated work still depends on a shop’s engineering and quality systems.

Should our shop list itself on Xometry or Thomasnet?

That is a business decision about margin and customer ownership. For AI visibility, complete and consistent directory profiles help because they are outside sources that confirm your capabilities and location.

Does our ITAR registration or CMMC status help in AI answers?

We have no evidence that it changes rankings. It matters because defense buyers filter on it. State it accurately and only if it is current.

Do engineers trust what ChatGPT says about suppliers?

Only partly. They rated trust in AI answers 4.7 out of 10 in the 2026 survey, so most will check your website before sending an RFQ.

Sources

Free strategy call

Some questions are easier to answer about your own business.

Bring the one that matters most. On a free 30-minute call we’ll take a first look at it and send you a short written read afterward.

Ask about my businessHow we work