Which Texas manufacturers ChatGPT names, and what their websites have in common
When a buyer asks ChatGPT for a control panel builder in Houston or a CNC shop near San Antonio, it answers with names and links. We wanted to know where those links lead and how the named manufacturers’ websites differ from the rest. In November 2026 we asked ChatGPT 39 buyer questions across five manufacturing segments and four Texas metros, then checked more than 500 manufacturers’ websites with the same detector.
Key findings:
- ChatGPT sends buyers to manufacturers’ own websites. 80% of the sources it cited (246 of 306) were the named companies’ own pages; directories were about one in five.
- The page it cites is usually a specific one: a capability page («UL 508A panel shop», «custom panels»), not the home page. 63% of named mentions were backed by a page on the company’s own site.
- Named manufacturers list their certifications far more often, in all five segments: 46–73% of their sites against 20–37% of the wider sample.
- For machine shops, capability detail stands out: a capabilities page on 49% of the named sites against 8%, an equipment list on 45% against 17%, stated tolerances on 23% against 5%.
- The quote path is weak almost everywhere. A quote form that asks for material, quantity and deadline appears on 2–11% of the sample’s sites, and on 5–19% of the named ones.
- Platform makes no difference: about half of both groups run WordPress.
What we asked and what ChatGPT answered
Two buyer questions per segment, asked for Houston, Dallas–Fort Worth, Austin and San Antonio:
- control panels and industrial automation;
- metal fabrication;
- CNC machining;
- plastics and rubber molding;
- electrical and industrial equipment.
A typical question: «I need a CNC machine shop in Houston, Texas for precision parts with tight tolerances. Which shops should I request quotes from?» Every answer named 5 to 16 companies, 229 different manufacturers in all.
What the answers cited:
| Source | Share of cited sources |
|---|---|
| The named manufacturers’ own pages | 80% |
| Directories and listings (Thomasnet, maps, LinkedIn and others) | about 18% |
| Other (news, associations) | the rest |
Houston answers leaned on the companies’ own sites most (78 own pages against 4 directory listings). Dallas–Fort Worth answers used directories more (59 against 20). Austin and San Antonio answers often sent the buyer to Houston firms or to shops 60–80 miles away.
How the named manufacturers’ websites differ
We ran the same detector over two groups:
- the websites of the manufacturers ChatGPT named (37–53 per segment);
- a wider sample of Texas manufacturers with 11–200 employees, by NAICS code (46–101 per segment, 320 in all).
Share of websites where the detector found each element (named / wider sample):
| Control panels | Metal fabrication | CNC machining | Plastics | Electrical | |
|---|---|---|---|---|---|
| Certifications named (ISO, AS9100, UL 508A…) | 73% / 20% | 47% / 37% | 64% / 34% | 46% / 36% | 51% / 26% |
| Capabilities page | 8% / 13% | 27% / 13% | 49% / 8% | 29% / 18% | 13% / 9% |
| Equipment list | 5% / 4% | 18% / 11% | 45% / 17% | 12% / 14% | 6% / 6% |
| Tolerances stated | 0% / 0% | 4% / 0% | 23% / 5% | 5% / 2% | 0% / 0% |
| Response time promised | 22% / 7% | 18% / 5% | 13% / 6% | 10% / 4% | 9% / 11% |
| Industries served | 16% / 13% | 18% / 8% | 19% / 14% | 54% / 30% | 34% / 24% |
| «Request a quote» link | 32% / 15% | 31% / 33% | 47% / 25% | 51% / 41% | 47% / 30% |
| Drawing upload | 8% / 7% | 16% / 10% | 21% / 10% | 29% / 11% | 17% / 15% |
| Form asks material, quantity, deadline | 5% / 2% | 6% / 11% | 19% / 8% | 15% / 9% | 13% / 9% |
What holds:
- Certifications are the steadiest difference: higher for the named group in every segment.
- Capabilities, equipment and tolerances matter clearly for machine shops, partly for metal and plastics, and hardly at all for panels and electrical equipment.
- A promised response time is more common among named panel builders, fabricators and machine shops.
- Platform, mobile layout and HTTPS don’t separate the groups: nearly every site has a mobile layout and HTTPS, and WordPress runs half or more of both.
What this means for a manufacturer’s website
These are correlations, not a formula. But they match what we see when we build manufacturers’ sites: buyers and AI assistants both look for the facts that let them shortlist you without a call.
- Name your certifications on the page, with the standard and the scope: ISO 9001, AS9100, UL 508A, ITAR.
- Give each capability its own page: the process, the materials, the size range, the tolerances you hold. That is the page ChatGPT cites.
- Publish your equipment list. For machine shops it’s one of the clearest differences.
- State the industries you serve and show a project for each.
- Say how fast you quote, and keep the promise.
- Make the quote form do the work: a drawing upload plus material, quantity and deadline, so your estimator can answer the first time.
- Check what AI actually reads: ask ChatGPT your buyers’ questions and see which of your pages it cites, if any.
Point by point for machine shops: CNC machine shop website checklist. Where quote requests get lost: manufacturing lead generation.
A word from the founder
«One step I’d give any manufacturer: make sure your website answers your buyer’s questions, from the certifications you hold to product specifications, prices and location, and talks about what interests your customer, not about how great you are. Manufacturers tell us AI assistants have started recommending them and they now get more requests from ChatGPT, Perplexity and others, so being named there has become their first goal. Inside their own companies, ChatGPT has become an everyday tool too.»
— Konstantin Tesov, founder of Uwindi
Method and limits
- Dates. ChatGPT questions and website checks on November 4, 2026.
- ChatGPT. Signed-out ChatGPT through an API, United States, English, no forced web search; one new chat per question. 40 questions, 39 returned data. Answers change from run to run; this is one snapshot.
- The wider sample. Texas manufacturers with 11–200 employees from a commercial company database, by six-digit NAICS code; corporations, chains, subsidiaries of foreign companies and non-manufacturers removed by hand.
- The detector. One script for every site: the home page plus up to four quote and capability pages. Shares are lower bounds: an element counts only if it’s found in the HTML, so forms in pop-ups or sites built in JavaScript can be missed. The same holds for both groups.
- Limits.
- The named and the wider groups are different sets of sites: only 1–5 named companies per segment are also in the sample, so we compare groups, not named against unnamed within one sample.
- ChatGPT’s list includes larger companies; company size may explain part of the gap, certifications especially.
- 37–53 named sites per segment: small numbers. Read the figures as direction, not precision.
- Names. We publish aggregate figures only and name no company, in either group.
To cite this study: «Uwindi, Which Texas Manufacturers ChatGPT Names, November 2026, uwindi.com/texas-manufacturers-chatgpt-study/». Journalists can request the segment tables: kt@uwindi.com.
Frequently asked questions
Where does ChatGPT get its recommendations for manufacturers?
In our 39 questions, 80% of the sources it cited were the manufacturers’ own websites, usually a specific capability page. Directories and listings made up about one in five.
What do the manufacturers ChatGPT names have on their websites?
Most often, named certifications: 46–73% of their sites list them, against 20–37% of the wider sample. For machine shops, capability pages, equipment lists and stated tolerances also stand out.
Does the website platform matter?
Not in our data. About half of both groups run WordPress.
Can a manufacturer make ChatGPT recommend it?
No one can guarantee that. What the data shows is that the named manufacturers publish more of the facts buyers check: certifications, capabilities, equipment and a clear quote path.
