New
Get your business discovered across ChatGPT, Gemini, Perplexity & Google AI Overviews — from $900/mo.
Get your business discovered across ChatGPT, Gemini, Perplexity & Google AI Overviews — from $900/mo.
Generative Engine Optimization (GEO) rebuilt this Miami metal supplier’s product and service pages into the source AI assistants cite for grade, thickness, and cutting-method questions.
C & R Metals
C & R Metals has supplied and distributed metal in Miami since 1985, serving Miami-Dade, Broward, Monroe, and Palm Beach counties from a 3,000-square-foot showroom on NW North River Drive. The company stocks more than 8,000 types and grades of aluminum, steel, and brass, plus over 4,000 showroom items including fasteners, hardware, abrasives, and paint. Alongside supply, C & R Metals provides waterjet, laser, and plasma cutting, machining, welding, fabrication, and coating services.
The Challenge
Demand in this market starts with technical reference questions: which grade and thickness suit a job, how cutting methods differ, whether an alloy fits a specific application. A growing share of those questions now go to AI assistants, which answer them without sending anyone to a website. C & R Metals had the expertise to answer all of them, but its product and service pages were built in a format language models could not extract from. Direct answers never opened a section, no question-and-answer structure existed, and category descriptions ran to a single general sentence with no technical parameters. The company’s real data — alloy grades, available formats, stock depth — appeared only in fragments, disconnected from the jobs customers select those materials for. AI answers on core industry queries were already being generated. They cited competitors.
We rebuilt a focused test group of pages around how language models select and quote sources. Weekly AI impressions grew 2.4× over the following three months, and the gain held rather than spiking and fading.
Choosing the pages: We built a test group of one page per key commercial intent, selecting for queries where AI features were already producing a detailed answer and crediting a competitor. Working from a limited set let us validate the approach and measure it cleanly before scaling across the rest of the site.
Restructuring the content: Every meaningful block now opens with a direct answer, and content is divided into short segments on a one-question-one-answer basis, phrased the way customers actually ask. Each topic was reduced to a self-contained passage a model can read without the surrounding page. We replaced general descriptions with verifiable data: alloy grades, available formats and sizes, and the technical characteristics of each processing method.
Markup and internal structure: We implemented schema markup so key entities — the organization, its services, its product range, and its service area — are read directly rather than inferred from prose. We then built an internal linking scheme that ties the test group into a single topical cluster with a defined role for every page, mapping product categories to the processing methods that apply to those materials. Each page now holds a ready answer to a core industry question in a form that can be quoted directly.
We measured AI impressions in Google's Generative AI features report from May 18 to August 15, 2026.
Weekly AI impressions grew from 471 in the opening week to 1,140 in the closing week — 2.4× growth
The pages accumulated 10.3K AI impressions across the measurement period
Growth began in late May, immediately after the changes went live
By mid-June the figure settled at 800–850 impressions per week and never returned to its starting level
Metal supply and processing has no meaningful seasonality, which rules out a seasonal explanation for the curve
The result rests on the structure of the content, not a one-time spike — and the same Generative Engine Optimization approach applies to the remaining product and service sections