AI Visibility for Manufacturers: Sourcing Prompts, Wrong Specs and Fixes

Industrial buyers use AI assistants to build a shortlist before they send an RFQ. Their prompts are specific. They combine a process, a material, a certification, a quantity and often a region.

Realistic examples (hypothetical, written for this page):

1. "Contract manufacturers in the Midwest that do 5-axis CNC machining of titanium with AS9100 certification."
2. "Which injection molders accept a minimum order of 5,000 parts for medical-grade polycarbonate?"
3. "ISO 13485 certified contract manufacturer for small-batch electronic assembly in Mexico."
4. "Sheet metal fabricators that can hold plus or minus 0.005 inch tolerances on stainless steel enclosures."
5. "Compare lead times for powder coating suppliers near Atlanta."
6. "Is [your company] ITAR registered, and what materials do they machine?"
7. "Alternatives to [competitor] for aluminum die casting under 10,000 units a year."

The buyer's words are not the words the engine searches with. OpenAI's help documentation says ChatGPT search typically rewrites a prompt into one or more targeted queries. So chasing exact prompt phrasing is a weak strategy. What holds up is having each fact (process, material, certification, tolerance, minimum order) stated plainly on a page an engine can fetch.

As of October 2026, five surfaces are worth checking for sourcing prompts: ChatGPT, Perplexity, Microsoft Copilot, Google AI Overviews and AI Mode, and Gemini. Claude matters for buyers who turn its web search on.

Each one draws on a different index:

• ChatGPT search depends on OpenAI's OAI-SearchBot. OpenAI's crawler page says sites opted out of it will not be shown in ChatGPT search answers.
• Google AI Overviews and AI Mode only link pages that are indexed and eligible for a snippet in Google Search, according to Google's AI features documentation.
• Copilot grounds on Bing's index. In Otterly's 2026 experiment, Copilot kept citing two sites after Google deindexed them, because Bing had kept them.
• Perplexity runs its own crawler, PerplexityBot.

Retrieval is not always on. When an engine answers without searching, it relies on training data frozen at a cutoff. OpenAI's model docs list a February 16, 2026 cutoff for the GPT-5.6 family and August 31, 2025 for the Instant path. A certification you earned this year won't be in those answers.

What this means: being healthy on one engine says nothing about the others. Our explainer on whether you need a Google ranking for ChatGPT covers the index split, and whether one dashboard can track every engine covers measurement.

Most manufacturer sites were built for humans who request a quote. That design hides the facts AI engines need.

Facts locked behind forms

Many shops put the capability list, material list or equipment list behind a "request our line card" form. No crawler fills in forms. If the only copy of your tolerances lives there, no engine can quote it.

Facts locked in PDFs

Spec sheets and certificates often exist only as PDFs. Google's crawling blog says Googlebot fetches PDFs up to 64MB. The evidence for how other AI crawlers handle PDFs is thin. Don't make a PDF the only place a fact appears.

Facts rendered by JavaScript

Part configurators, material selectors and tabbed capability pages often load content with JavaScript. Vercel's crawler study found ChatGPT and Claude crawlers fetch JavaScript files but don't execute them. Googlebot renders; most AI crawlers do not.

Facts that contradict each other

Say you run a hypothetical machine shop. Your homepage says "ISO 9001 certified," your quality page lists AS9100, and an old press release mentions a minimum order of 500 units that you dropped two years ago. An engine that reads all three can repeat any of them.

The move: give every load-bearing fact one current, plain-text statement in the HTML of a page that loads without a form or a script.

These are the facts that decide whether you make a buyer's shortlist. Audit each one against what AI engines say and against what your own pages say.

Certification logos shown as images carry no text a crawler can quote. A specific sentence ("We hold AS9100D certification for CNC machining at our Dayton plant") gives an engine something it can repeat. Stating specific facts this way is the only on-page approach with positive experimental signal in the evidence. It is not a guarantee of citation.

Sourcing answers often cite directories and lists rather than manufacturer sites. When engines cite other sites instead of yours, being named accurately on those sites is the route into the answer. That is an association in the evidence, not a promised effect.

Check these, in this order:

• Industrial supplier directories. Confirm your processes, materials, certifications and minimum orders match your site today.
• Certification body registries. Make sure your certificate shows as current under your legal name.
• Trade association member directories. These often hold years-old descriptions.
• Distributor and partner pages that describe what you make.
• B2B marketplaces where you list parts or services. Our guide to AI visibility for online marketplaces covers how those listings surface.
• Your LinkedIn company page and recent trade press, especially after an acquisition or plant closure.

The next move: when an AI answer gets a fact wrong, look at the sources it cites. The wrong fact usually sits in one of them, and that listing is what to correct.

Content fixes are worthless if crawlers can't reach the page. Run these checks first.

1. robots.txt, per bot. Rules are independent. OpenAI's crawler page says each setting is independent: GPTBot governs training, OAI-SearchBot governs ChatGPT search. A blanket AI block often removes search eligibility by accident. A minimal allow looks like this:

User-agent: OAI-SearchBot
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: Claude-SearchBot
Allow: /

2. Firewall and CDN. A firewall or CDN can block an AI crawler that robots.txt allows, often with a 403. Manufacturers with strict security setups are exposed here. From September 15, 2026, Cloudflare blocks Training and Agent crawlers by default on ad-displaying pages for new domains and existing free-tier customers, while Search stays allowed. Check your settings even if you never changed them. The AI Crawler Checker requests a page as each AI crawler and shows which ones get turned away.

3. Rendering. View your capabilities page with JavaScript disabled. If processes or materials disappear, most AI crawlers can't see them either.

4. Snippet and index controls. Look for noindex, nosnippet, max-snippet:0 and data-nosnippet copied from a template. Google and Bing document these as controls that keep content out of AI answers.

5. Google Search Console. Inclusion in Google's generative AI features is on by default, but the opt-out control can be set, and child properties inherit it. Confirm it.

Days 1 to 7: baseline. Write ten sourcing prompts buyers would use to find you. Run them on each engine, logged out, and record whether you're named and which facts are wrong. Our guide on how to check brand visibility in AI search walks through the method. The free ChatGPT brand mention check gives a quick snapshot to start from.

Days 8 to 14: access. Fix robots.txt per bot, test the firewall for each crawler, and remove stray snippet controls. Check Search Console and Bing Webmaster Tools separately.

Days 15 to 21: facts in HTML. Publish one capabilities page with processes, materials, tolerances, certifications and minimum orders in plain text. Keep the PDFs as downloads, not as the only copy. Remove contradictions across old pages.

Days 22 to 30: third parties and recheck. Update directory, registry and association listings. Rerun the same prompts and compare. To track change over months instead of once, follow our steps for measuring AI share of voice.

Expect access fixes to show up only after engines recrawl. No one can promise a citation, but every step above removes a reason to be left out.