More buyers now start a supplier search on ChatGPT, Perplexity or Google’s AI answers instead of scrolling through ten blue links. They ask a question in plain language and get a shortlist back. If your company is not in the material those systems read, you are not in the conversation — even if you never lost a single order to a competitor.
This page answers the questions buyers and suppliers ask about how AI systems find fabrication suppliers, and what actually decides whether a shop shows up.
How AI Search Actually Finds Fabrication Suppliers
The short answer: an AI assistant does not crawl the web live in most cases. It answers from an index built earlier — by its own crawler, or by a search engine it is connected to. So “getting found by AI” reduces to three things: can the crawler reach your pages, does it read your content, and does your content answer the question being asked.
That means the usual SEO work still matters. What changes is the format and the clarity that get rewarded.
Questions Buyers Ask About AI Visibility
Q: Do I need a special “AI SEO” service?
No. The foundations are the same as conventional search: crawlable site, indexable pages, real content, clear structure. What differs is that AI answers favour pages that answer a specific question directly, in the first sentence or two, rather than pages that build up to it.
Q: Does my robots.txt block AI crawlers?
This is the first thing to check, and the most common own goal. Most sites set User-agent: * and allow everything — which is fine. Trouble starts when a host, a plugin, or a security setting silently blocks specific crawler user agents. That block is invisible in the browser (you see your site normally) but total for the crawler.
The practical test: request your own pages while sending an AI crawler’s user agent string, and compare the status code with a normal browser request. If the browser gets 200 and the crawler gets 403 or 429, the crawler is being blocked — and no amount of content will help.
Q: Does my site need JavaScript for the content to be read?
It should not. Many crawlers do not execute JavaScript, or execute it in a limited way. If your product descriptions, specifications and articles only appear after scripts run, an AI system may see an empty page. Server-rendered HTML that already contains the text is the safe default.
Q: What makes content more likely to be quoted?
Specific, structured, and self-contained. A paragraph that names the process, the material and the constraint in one place is easier to lift into an answer than a page that spreads the same information across five sections. Explicit question-and-answer sections, comparison tables and short definition blocks all help, because they map directly onto how a question is phrased.
Q: Should I write for AI or for people?
For people — but write in a form that survives being summarised. That means stating the conclusion early, keeping each section about one thing, and avoiding marketing language that carries no information. AI systems summarise what is clear; vague copy summarises into nothing.
Q: Do I need a llms.txt file?
It is a proposed convention, not a standard, and no major AI system has committed to relying on it. It can help orient a system that chooses to read it, but it is not a substitute for crawlable pages. A site with perfect HTML and no llms.txt will do better than a site with a llms.txt file and pages that crawlers cannot reach.
Q: How do I know whether AI systems are reading my site?
Check server access logs for the crawler user agents — GPTBot, ClaudeBot, PerplexityBot, OAI-SearchBot and similar. If they never appear, the crawler is either not interested or being blocked. If they appear but return 4xx, they are being blocked. Neither shows up in normal web analytics, which is why this check is usually skipped.
Q: Does this replace search engine optimisation?
No. The overlap is large. A page that ranks well in ordinary search is usually the same page an AI system draws from, because most AI answers are grounded in search results. Fixing crawlability, structure and clarity helps both at once.
What Actually Decides Whether a Supplier Shows Up
- The crawler can reach the page. No robots block, no user-agent block, no rate limit that trips on a crawl.
- The page contains the answer in text. Not in an image, not behind a script, not only in a PDF.
- The page is specific about capability. “Heavy steel fabrication” describes an industry. “Plate to 300 mm, welded frames machined after welding, 8,800 m² workshop” describes a supplier an answer can point to.
- The page covers the question the buyer actually typed. Buyers rarely search for a process name alone; they search for the problem, the constraint, or the comparison.
- The site is consistent about who it is. Names, locations and contact details that agree across pages are easier for a system to cite with confidence.
Where Suppliers Usually Lose Visibility
In practice the failures are rarely about content quality. They are mechanical:
- A silent crawler block. A security plugin, a CDN rule or a host setting rejects a user agent, and nobody notices because browsers are unaffected.
- Content that never reaches the HTML. Pages built entirely client-side look complete in a browser and empty to a crawler.
- One page trying to answer everything. A single long “services” page is harder to match to a specific question than several pages that each answer one.
- No plain-language statement of capability. Specifications buried in PDFs or brochures do not get read.
- Soft 404s. Pages that return “200 OK” but contain an error message confuse every system that tries to build an index.
How VHOFAB Approaches This
VHOFAB publishes its capabilities, materials, tolerances and inspection approach as server-rendered pages, so the text is present in the HTML a crawler receives rather than assembled by scripts afterwards. Content is organised by question — what to specify, how a process behaves, what a buyer should check — rather than by service name, and each page states its subject directly.
If you are assessing a supplier on how findable they are, the checks above apply to us as well: crawler access, server-rendered text, specific capability statements, and pages that answer one question at a time. To discuss a component, send the drawing and technical requirements through the contact form.