Why Manufacturers Don't Show Up in AI Search
Tomohiro Iida · Published July 11, 2026
"Who can I ask for precision machining of X?" is a question procurement staff are now putting to ChatGPT or Perplexity. Whether your company's name comes up in the answer depends not just on your technical skill, but heavily on whether that technical information is published on the web in a form AI can read. This article covers the manufacturing-specific reasons manufacturing sites get missed in AI-search answers, and the steps for moving away from catalog-PDF dependence toward technical information AI can actually check.
Key takeaways
- A PDF itself is something Google Search can index. The problem isn't 'having a PDF' — it's that key information is locked inside PDFs only, without product- or technology-specific HTML pages, internal links, calls to action, or measurement. Scanned-image PDFs need extra checking.
- The core fix is turning the information procurement staff use to shortlist vendors into text, structured on web pages. Not everything needs to be public — design what you publish versus what you keep for negotiation.
- No one can guarantee inclusion in an AI answer. What you can do is prepare source material AI can cite, and track changes through repeated, consistent test questions.
The real problem: unreadable and unanswered
A manufacturing site's weakness isn't the existence of PDFs — it's that the specs, materials, tolerances, lot sizes, equipment, and inspection information procurement staff need aren't available on product- or technology-specific HTML pages.
AI search means an experience where ChatGPT Search, Perplexity, Google AI Overviews, and similar tools generate the answer itself and cite the web pages it drew on as sources. The material for that answer is text on a web page, and AI tends to pull passages — a heading plus the paragraph, table, or list right after it — rather than reading a whole page.
- Materials, tolerances, and equipment information exist only inside a catalog PDF
- The information needed to judge feasibility sits only behind a "please contact us" prompt
- A company-profile page exists, but nowhere states in writing what, under what conditions, and how far the company can go
This is a problem of how information is placed, not of technical capability — which means it's fixable by changing the placement.
Why procurement research now looks like an AI question
AI search matters to manufacturers because the early-stage "finding candidates" questions procurement and purchasing staff ask now take the same shape as a question to an AI.
- "Who can I ask for precision machining of X?"
- "Which makers can produce part Y in small lots?"
- "Can you point me to a machining company that handles this material and this tolerance?"
- "Is there a company that will take on prototyping from a single unit?"
These used to be typed as keywords into a search engine. When the same question goes to ChatGPT or Perplexity instead, the AI assembles its answer through the following flow: it interprets the question and runs a web search; it extracts the passage that answers the question from the search results; and it generates an answer grounded in that extract, attaching the source URL.
Manufacturing sites tend to stumble at each stage of that flow. If a page isn't crawled or indexed, it can't become a candidate in the first place; if key information sits only in a long PDF, it's harder to design product-specific search intent, internal links, and measurement around it; and without text that directly answers the question, there's no clear passage to cite as grounds for an answer.
General, industry-agnostic reasons content goes uncited — exaggerated claims, inconsistent facts, unsupported statements — are the focus of a separate article; this one stays focused on patterns specific to manufacturing.
Five reasons manufacturing sites are not cited
The manufacturing-specific reasons are: catalog-PDF dependence, an information structure limited to "contact us," no direct answers to procurement questions, inconsistent terminology and missing structured data, and unconfirmed crawler settings.
- Reason 1: catalog PDFs and drawing images are the primary source of information
- Product specs, supported processes, and equipment lists exist only inside PDFs or images, with no product- or technology-specific HTML page. Google can index PDFs, but a PDF alone weakens page-level internal linking, calls to action, update management, analytics, and responsiveness to product-specific search intent. Always verify text extraction and passage location on the actual file for scanned-image-only PDFs or long catalogs covering multiple products. The fix isn't to eliminate PDFs — it's a division of labor: turn your main products and processing technologies into text as web pages, and keep the PDF as a supporting resource (detailed drawings, the full catalog).
- Reason 2: key information stays behind a "please contact us for details"
- Deliberately keeping capability, precision, and lot-size details private and routing everyone to an enquiry form. That works for human sales conversations, but in an AI-search context it reads as a site with no citable information. AI can't retrieve what's behind a contact form, so it tends to be left out of answers built from public information alone. The fix is to separate the information used for shortlisting from the information used in negotiation, and publish the former.
- Reason 3: no page directly answers procurement questions
- A company profile, history, and message from the president may be well developed, while there's no page that answers, in writing, whether the company can machine a given material, what its minimum lot size is, or whether it can do prototyping. Because AI tends to cite at the passage level, without a heading and paragraph matching the question there's nothing to extract, however good the overall impression of the site is. The fix is an FAQ or technical-information page built around procurement questions as headings, with the conclusion stated right after.
- Reason 4: inconsistent terminology and missing structured data
- When company name, product name, equipment name, and material name vary from page to page, it's harder for both people and search systems to confirm they refer to the same thing. Use Organization, Product, and BreadcrumbList structured data where it fits the page, matched to the body text. That said, Google has said no special schema is needed for its AI features, so structured data alone is not a guaranteed way to increase AI citation.
- Reason 5: AI crawler permissions have not been checked
- robots.txt and CDN settings haven't been reviewed in years, unintentionally blocking crawlers used for search. At OpenAI, OAI-SearchBot is involved in ChatGPT Search inclusion while GPTBot is for model-training use, and the two can be controlled independently. Google's AI Overviews and AI Mode draw on Googlebot's search index, so Google-Extended settings should not be conflated with search visibility.
Five steps to build citable technical content
Work through the steps in order: inventory questions, turn them into pages, design disclosure scope, add structure, then confirm crawler access and start monitoring. Implementing structured data alone, skipping the upstream work, won't help if there is nothing worth extracting.
- STEP 1: Inventory procurement questions
- Gather procurement-side questions from past enquiry emails, phone notes, trade-show questions, and a conversation with sales about what people usually ask. Sorting them by material, tolerance/precision, dimension range, lot size (min/max), prototyping feasibility, lead time, inspection process, and certifications (ISO, etc.) gives you a structure you can use directly for page content.
- STEP 2: Turn main products and processes into web pages
- Build a page per process or product that answers the inventoried questions. Phrase headings as questions or in a form where the answer is anticipated; state the conclusion in two to three sentences right after the heading (answer-first); and present materials, precision, dimensions, and lot sizes in both prose and a table. Reposition the PDF catalog as a supporting resource at the end of the page.
- STEP 3: Design disclosure scope
- Not everything needs publishing. Split information into what you publish and what you keep for negotiation, and put full effort into turning the published side into text.
- STEP 4: Implement consistent terminology and structured data
- Align company name, location, equipment names, and material terminology across every page, and add Organization, Product, BreadcrumbList, and other structured data that fits the page content, matched to the body text. Keep FAQ content as visible HTML, but don't implement FAQPage schema aimed at Google's FAQ rich result. Don't write specs, certifications, or track record into the schema that aren't also in the body text.
- STEP 5: Confirm crawler access and start ongoing monitoring
- Decide your policy for major AI crawlers in robots.txt, and confirm the pages you want cited are actually readable. Then put representative questions from STEP 1 to AI regularly under the same conditions, and record how mentions and citations of your company change.
A standard template for product and technology pages
Structure your main product and processing-technology pages in the following order. This can be copied or printed as an in-house content checklist.
| # | Section | What to include |
|---|---|---|
| 1 | Conclusion: what processes/products you can handle | State up front, at the top of the page, what you can handle and how far |
| 2 | Materials, dimensions, precision/tolerance, lot size, and typical lead time | Show procurement's decision-making inputs in both prose and a table |
| 3 | Equipment, inspection process, and certifications | State the equipment you hold, your inspection process, and ISO or other certification status concretely |
| 4 | Conditions you can't accommodate, or need to discuss | State materials, sizes, or quantities you do not handle, and the conditions under which you will consider a request |
| 5 | Track record and use cases (within what you are cleared to publish) | Cite case studies or representative applications you have permission to publish |
| 6 | Links to drawings and PDF catalogs | Put detailed drawings and the full catalog as a supporting resource at the end |
| 7 | Information needed for a quote, and the contact path | State what to include in a quote request and where to send it |
| 8 | Last-updated date, technical reviewer, and revision history | Show how current the information is and who is accountable for it, to support credibility |
How much to disclose: answering the competitive-exposure worry
The disclosure rule is: publish what procurement uses to shortlist vendors, and keep the pricing breakdown and process know-how — the real competitive core — for negotiation. Publishing technical information naturally raises the worry that competitors will see your playbook, but that worry can be addressed by designing the level of disclosure.
| Category | Examples | Handling |
|---|---|---|
| Publish (used for shortlisting) | Supported materials, processes, precision/tolerance range, dimension range, lot-size handling, prototyping feasibility, main equipment, certification status, typical lead time | Turn into text and structure it as web pages |
| Keep for negotiation (used in the award decision) | Individual quotes, pricing breakdowns, detailed drawings, process-design or jig know-how, terms with specific customers | Disclose after an enquiry or under NDA |
When you are unsure about a piece of information, judge it by whether, without it, a procurement team could even shortlist you. If information needed to be shortlisted stays closed off, the negotiation opportunity never arises in the first place. Conversely, there is no need to publish the process know-how that is your actual competitive edge.
Measuring impact, and its limits
No company can guarantee inclusion in an AI-search answer. What you can confirm is change, through repeated observation under the same question and the same conditions.
What each AI engine chooses to cite is that engine's own call, and answers vary day to day. This isn't a case of "fix it and you're guaranteed a recommendation," so impact is measured by putting representative procurement questions to ChatGPT, Perplexity, and similar tools on a regular schedule under the same conditions and recording whether your company is mentioned, which URLs are cited, and what is said; by checking search- and AI-referred traffic and enquiries in your analytics; and by checking, at the same time, whether any inaccurate description — an old company name, a discontinued business, a wrong address — is showing up.
Improvement isn't a one-time fix — it's a repeating loop of building, observing, and correcting.
FAQ
- Is having a catalog PDF on our site enough?
- Whether a PDF alone is enough depends on the content and the goal. Google can index PDFs, but also publishing your main product and technical information as HTML makes it easier to design around product-specific search intent, internal links, updates, calls to action, and measurement. A practical division of labor is to keep PDFs as drawings, spec sheets, and catalogs, and use HTML as the entry point for shortlisting.
- Won't publishing technical information let competitors copy us?
- What you would publish is the information procurement uses to shortlist vendors — supported materials, precision, lot sizes, and similar — while process design, jigs, and pricing breakdowns, the real core of your competitiveness, stay in negotiation. Designing and separating what you disclose lets findability and confidentiality coexist.
- Can a small manufacturer without a dedicated web person do this?
- Yes. The first step is not web expertise — it is inventorying frequently asked questions, which can be done with sales and shop-floor knowledge alone. Building the pages and implementing structured data can be split off to an outside partner. To check where you currently stand, a free scan can assess how your company appears in AI search from your public site URL alone.