Absent from Google's AI, then first: a B2B industrial equipment manufacturer
An Indian manufacturer of engineered equipment for cement, lime and metals plants had decades of plant experience and a website three weeks old. ChatGPT already recommended it on some buyer questions. Google's AI Overviews named competitors on every one, and credited one of them with the client's own signature claim. Geology ran a GEO engagement: a technical remediation sprint, a settled company identity, 41 new articles written for plant engineers, and six off-site placements. Google's AI Overviews now name the client on 50%+ of buyer questions, up from none, and both engines name it first on the plain category search.
About B2B industrial equipment manufacturer
A B2B industrial equipment manufacturer in India, supplying engineered components for heavy process equipment in cement, lime and metals plants, with buyers in India and Europe. The website runs in English, Italian and German.
The buyers are process and maintenance engineers with capex budgets. They plan purchases around a scheduled plant shutdown, months in advance, and they research the problem long before they contact a supplier.
The Challenge
On day one the site had 34 pages, three blog posts and no visible backlink profile. The two AI engines disagreed about whether the company existed. ChatGPT read the new site and recommended the client on five of nine buyer questions in our audit. Google's AI Overviews recommended it on none, and on one question credited a competitor with a claim the client makes about its own product.
The brief was short: be the supplier AI names. That meant a remediation sprint first, then a retainer for content, off-site authority and AI tracking. Every page had to hold up when a plant engineer read it.
The plan rested on one observation. Engineers search the symptom (fuel use, heat loss, shutdown time) well before they search for a supplier. A site that answers those questions with checkable figures gets read by the people who sign off the purchase, and gives AI engines sentences they can quote and attribute.
Our approach
A remediation sprint before any new content
The site was three weeks old, in English, Italian and German, and Google could not tell the language versions apart. We fixed metadata and canonicals, added hreflang for every page, rebuilt the sitemap, added structured data where there was none, and cleared the Core Web Vitals failures. New content would have been wasted on a site Google read as three copies of itself.
Settle who the company is
Several unrelated businesses share the client's first name, and AI engines blended them. We rewrote the About page and the structured data to state the trading name, the parent company and the legal entity in plain sentences, so an engine reading any page could tell which company it was looking at. Google's AI Overviews had credited the client's signature installation claim to a competitor, and fixing the identity was the first step to getting that claim back.
From 3 articles to 44, written for plant engineers
Process and maintenance engineers search the symptom before they search for a supplier: heat loss, fuel consumption, air leaks, wear. We planned 41 new articles in three layers. Problem pages explain how to measure each symptom and what it costs a plant. Application pages cover each industry the client serves, so an engineer in lime or metals finds their own process. Buyer pages compare designs and explain retrofits, spare parts and replacement intervals.
An editorial standard an assistant can quote
Every page opens with a direct answer, then names the standards, figures and plant conditions behind it. An engineer can check each claim, and an AI assistant can lift a sentence and attribute it to the client.
Six placements and 17 links, from zero
The site started with no visible backlink profile. We placed six guest articles in business and industry publications, each making a specific claim about the client that a reader can check. One of them is now the source Google's AI Overview cites when it names the client first.
Tracking what the engines say
We ran a fixed set of non-branded buyer questions through ChatGPT and Google's AI Overviews and recorded who each answer named, from the pre-engagement audit to the final tracked run. That tracking is how we spotted the misattributed claim, and how we know which pages the engines quote.
Results
AI figures compare our pre-engagement audit with the final tracked run, on non-branded buyer questions only. Search figures compare the first four weeks of Google Search Console data in the engagement with the latest four weeks (to mid-November 2026), on the client's verified domain property. All multiples are rounded down.
The client paused the work after July. We measured "now" after the pause, not at the peak, and search demand has eased since then: impressions in the latest four weeks sit well below the summer high. The multiples above are what remains three months into the pause.
The AI positions have held. On the plain category search, with no country named, ChatGPT lists the client first of five suppliers and tags it as the Indian specialist to evaluate. Google's AI Overview names it as the first manufacturer, ahead of the competitor it once credited with the client's claim, and cites one of the off-site articles Geology placed. On narrower buyer questions, such as the best manufacturer in India or a specific product design, both engines again put the client first.
Selling equipment to engineers?
Geology specializes in insurance, financial services and B2B. A free audit shows how ChatGPT and Google's AI Overviews answer your buyers' questions today, and which pages they would need to cite you.