Mechanical engineering
Process, material, envelope, tolerance, volume: the buyer's question is a specification.
Before an RFQ goes out, an engineer may ask ChatGPT or Perplexity which shops can machine a titanium housing of [size] to [tolerance], [quantity] a year, within reach of the plant. With web search on, the assistant reads a handful of pages and can reply with a few supplier names and a reason beside each.
A supplier question names a process, a material, a part size, a tolerance, a quantity, an industry and sometimes a certificate. Those are the figures an answer would have to repeat about your firm, and your own site is where you state them all: machines and envelopes, tolerances held, materials, part sizes, volumes, industries served, certifications held and lead times, each with a unit and a date.
How your customers ask
Typed before an RFQ goes out
A supplier question carries the drawing's numbers
The first question lists what a drawing and a purchase order would: a process, a material, a size limit, an annual quantity, a place. The others add a cycle time, a weight, a certificate, an industry and a service territory. Each is a condition a supplier meets or does not, and nearly every one has a unit.
With web search on, an AI assistant reads some of the pages its searches return before replying; it can also answer from its training. Either way, a reply can state your tolerance correctly only if some text states it. "High precision" gives it no figure to repeat, and a datasheet revised years ago gives it an old one.
Some conditions turn on rules rather than on your shop: machinery safety and conformity marking, export controls, public procurement. Those differ between the US, Canada, the EU and the UAE. Your pages state what the firm builds, holds and has documented; what a rule requires is for your counsel or the authority where the machine is sold.
Where each specification belongs on the site
| The buyer asks | What the firm states | Where it goes |
|---|---|---|
| Can you do this process? | Each process by name, and the machines behind it by type, axes and count | A capabilities page, one section per process |
| How big can the part be? | Each machine's work envelope (X, Y and Z travel, turning diameter and length, maximum part weight) in millimeters and inches | A machine list in HTML, linked from each process |
| How tight? | The tolerance held routinely and the one offered on request, per process and feature, with how it is inspected | Beside each process |
| Which materials? | The materials and grades you cut, form or weld, and the ones you do not take | The capabilities page, as a list |
| How many? | Batch sizes and annual volumes, from prototype to series, as [n] to [n] parts | The capabilities page and the RFQ page |
| Which industries? | The industries served, with the kinds of parts or machines made for each | A page per industry you want to be asked about |
| Which certificates? | Each certificate held: issuer, number, scope in the certificate's words, valid from and until, where to verify it | A quality page, linked from each process it covers |
| How long? | Typical lead times by job type: quote in [n] days, first article in [n] weeks | The RFQ page |
| What does the machine do? | A datasheet per standard machine or component as an HTML page, with the PDF as a download beside it | A page per machine or product line |
| Which company is this? | One legal name, each plant's address, one main number and one RFQ address, the same everywhere | Every page, profile and directory entry |
Three habits keep the table honest: a unit and the month last checked on every figure; tolerances written per process and feature, the way your inspection proves them; and each certificate stated as issued, no wider than its scope. Capability pages a buyer can compare works through the tables row by row.
Organization for the firm, LocalBusiness for the plant people visit
We read schema.org on 2026-09-25. It has no type for a machine builder, a machine shop or a manufacturer, and it marks ProfessionalService deprecated. Organization describes the company. LocalBusiness, "a particular physical business or branch of an organization", sits under both Organization and Place and takes Place properties such as opening hours, so it suits a plant customers visit for an acceptance test or a pickup, with parentOrganization pointing to the company.
Under the company, each process you sell can be a Service (serviceType, provider, areaServed, category), and each standard machine or component a Product, whose properties include material, width, height, depth, weight, model and additionalProperty for a figure with no property of its own. Certificates go in hasCertification as a Certification: issuedBy, certificationIdentification, certificationStatus, validFrom, expires and validIn, copied from the certificate.
Write the structured data as JSON-LD, which Google's introduction to structured data recommends; the same page says not to mark up what a visitor cannot see, so each node repeats a table on its page. Leave ratings off the company's node: under Google's review snippet rules, an organization that controls the reviews about itself is ineligible for review stars.
Our crawl's figures, read for an engineering supplier's site
Our audit of 9,483 small business websites fetched each site's homepage, contact page and about page with a crawler that ran no JavaScript and obeyed robots.txt, in 40 trades and 90 metros across the US, Canada, Europe and the UAE, in late August and early September 2026. The sample was drawn per metro and trade, not from a random frame. Every figure below is the whole sample's, across all 40 trades; we publish no engineering-only cut.
2.1% kept their real content, such as the price list or the service menu, in a PDF, and 3.0% served their text only through JavaScript. On an engineering site the same gap would be a datasheet kept only as a PDF, or a configurator that draws its figures with a script. 7.2% served our crawler a challenge page, and 3.2% disallowed AI crawlers in robots.txt on purpose.
61.5% had no question-and-answer content of any kind, and 50.7% had a question as a heading with nothing under it that a machine could isolate. 43.1% were inconsistent about who and where they are across those three pages: 36.9% showed more than one phone number, and 34.1% had no postal address a machine could parse. 17.8% had no schema.org markup, and 13.7% had business markup missing the fields that would make it usable.
The probe beside the audit put one local trade question per business to Claude, with web retrieval on, on one day: 14.9% of 9,240 businesses were named. It is a figure for the whole sample, from one question on one day, and no rate for engineering suppliers.
The order to work in, from the served page to the first reading
- Run the crawler access check, then open a capability page with scripts off and find one tolerance and one envelope in the text. If your CDN serves crawlers a challenge page, allow the answer-side AI crawler tokens by name.
- Make the identity one thing: the legal name, each plant's address, one main number and one RFQ address, the same on every page, profile and directory entry.
- Rebuild each capability page around dated tables with units, as capability pages a buyer can compare lays out.
- Turn each datasheet into an HTML page with the PDF as a download beside it; the PDF trap covers what a fetcher gets from a file.
- Publish the quality page: each certificate with its issuer, number, scope and dates, and no claim beyond it.
- If you lead a narrow niche, publish the evidence: applications solved, the installed base by region, the standards your machines are built to (a narrow lead, stated as evidence).
- Add Organization, Service and Product markup, with LocalBusiness for a plant visitors come to, repeating the tables and adding nothing.
- Take a first reading with bracketed capability questions (named for the part you can make), then repeat it monthly beside your inquiry log (the routine for a long sales cycle).
These steps change what your pages state and what a crawler can fetch, not what an assistant's operator does with them. With monitoring, the monthly reading under our methodology shows whether the answers moved.
Go deeper
The articles behind each step
Fundamentals
Named for the part you can make: AI answers for machine builders and shops
Technical & Structure
Capability pages a buyer can compare: specifications in dated HTML tables
Content & Answer Pages
The PDF trap: when your price list is a file, an assistant cannot quote it
Brand & Positioning
A narrow lead in machine building, stated as evidence a buyer can check
Measurement & Reporting
Monthly readings of AI answers for an engineering firm's long sales cycle
Questions from the shop floor and the sales office
Our customers' parts are under NDA. What can a capability page show?
The capability, not the part. Envelopes, tolerances by process, materials, volumes and industries describe what your shop can do without describing anyone's drawing. Write an application up by industry and region, and name the customer only with written permission. Photos of a customer's part need that permission too; photos of your own machines do not.
Lead times depend on the backlog. Should we publish any?
Publish typical ranges by job type, dated, with what moves them: "quote in [n] working days; first article in [n] to [n] weeks from an approved drawing, as of [month year]". A dated range tells a buyer what to plan around. Without one, an answer has only what another page says about your lead times.
We build one-off machines. What goes on a product page?
A page per machine family or application rather than per machine: the task, the parts and materials it handles, the rates you have built for, footprint, utilities and the standards it is built to, each as a bracketed figure or the standard's own name. Link the applications delivered in that family, anonymized where the customer asks.
Our work comes from repeat customers and trade-show contacts. Is any of this for us?
The part that starts with a stranger is. A new engineer at a customer, a buyer qualifying a second source, or someone you met at a show may look you up before writing. A capability page with figures answers the stranger and confirms what the visitor heard at the booth.