How a 30-Year Foundry Got Into Google's AI Overview

Kansas Castings is a 30-year-old American iron foundry that pours specialty high-chrome white iron. The product was excellent. The website wasn't. It was thin and unranked, there was no inbound pipeline, and the company was effectively invisible for the exact searches its buyers run, phrases like "white iron castings supplier."

Within a week of relaunching the site, Kansas Castings showed up in Google's AI Overview for a high-intent buyer query, alongside much larger national foundries. Since then the site has brought in a real inbound request for quote, still in process, along with a small bump in orders and more calls to the owner asking for quotes. The whole effort, from building it to putting it to work, has taken about six months, and it's still going.

None of it required a big budget or an agency. Here's what moved the needle, in order.

1. Write for the searches buyers actually run

Most small-manufacturer websites describe the company. Buyers don't search for the company. They search for the part, the material, or the problem.

I rewrote titles, meta descriptions, and on-page copy across all 11 pages around what buyers type, and replaced filler with accurate, buyer-intent content. I also found a site-wide missing-H1 problem and fixed it once, at the template level, instead of page by page. I did the rebuild through the WordPress REST API with Python scripts, so changes were consistent and repeatable.

The buyer phrases came from three places: the owner, who knows how his customers describe what they need; Google Search Console, which shows what people actually typed before landing on the site; and competitor sites, which show what the market already ranks for.

If you're doing this: start with five or ten phrases a real buyer would type. Ask the people who take the sales calls, then check them against Search Console. Make sure each phrase has a page that answers it plainly.

2. Tell search engines exactly what you make

Search engines, and the AI answer engines built on top of them, read structured data before they read your prose. Most small manufacturers have none.

I added a full JSON-LD schema stack: Organization, LocalBusiness, Product, and FAQ markup. That told Google, in a format it trusts, what the foundry makes, where it is, and who it serves. That layer is what made the AI Overview placement possible.

If you're doing this: Product and FAQ markup do the most for a manufacturer. Answer the questions buyers ask your sales team, on the page, in FAQ markup.

3. Make sure a quote request actually arrives

A lead form that silently drops submissions is worse than no form, because you never know what you lost.

I built a custom RFQ quote form with server-side handling, email routing, and a database backup of every submission. Getting it to deliver meant debugging a hosting firewall quirk and SMTP authentication. Every quote now lands in the owner's inbox, and a copy is kept even if email fails.

If you're doing this: submit a test quote yourself every month. Keep a copy of every submission somewhere other than email.

4. Don't wait for inbound

SEO compounds, but slowly. Outbound fills the gap, as long as it doesn't damage the company's main email reputation.

I set up a separate sending domain on Google Workspace with DKIM, SPF, and DMARC authentication and a warm-up schedule. Prospects came from Apollo.io, targeted by industry vertical, and Smartlead ran multi-step sequences. In deliverability testing, it reached 100% inbox placement.

If you're doing this: never send cold email from your main domain. Authenticate the sending domain and warm it up before you send a single sequence.

5. Measure from day one

GA4 and Google Search Console went in with the relaunch, so we could see which searches the site showed up for and which pages earned clicks, instead of guessing.

How one person did all of this

I'm not an agency, and this wasn't a team effort. My main tool was Claude, an AI assistant, which did much of the writing, scripting, and debugging with me. Alongside it I ran OpenClaw, an AI agent gateway on my own server, working through Codex. Together they let one person with an operations background do the work of a web developer, an SEO specialist, and an email deliverability engineer.

The short version

1. Write each page for a search a real buyer runs.
2. Add structured data so search and AI engines understand what you make.
3. Make sure every quote request arrives, and keep a backup copy.
4. Run outbound from a separate, authenticated, warmed-up domain.
5. Measure with Search Console from day one.

The full project is in the Kansas Castings case study. If you run a small manufacturer that's invisible online, I'm glad to compare notes.