Case study, India
A marble, granite and natural stone company. We built the website, product architecture, content, technical SEO, generative engine work and authority from the ground up.
One client, one market, one six month window. Not a forecast for your site.
This page follows the work in order: the problem, what we built, how we measured it, the numbers as reported, the limits of what they prove and what we would do differently. Every figure is taken from a report we can show, and the screenshots further down are those reports.
Where this page describes method in general terms, it says so. The honest summary is that this is a story about a trade with many strong suppliers and few well built websites. The opportunity came from that gap as much as from the work itself.
Evidence
Six months. Sources are named in the measurement section below.
One client, one market, one six month window. Not a forecast for your site.
Starting point
Premium photos, weak headings and unclear categories.
Little on price, quality, origin or comparison.
Inconsistent facts and few authoritative mentions.
Posts that do not compound like a knowledge base.
Attention wasted by poor pages and tracking.
Rank reports could not show AI mentions.
The problem
Sudarshan Stoneix deals in marble, granite and natural stone. The products were good and the photography was premium. The website looked the part. What it did not do was help a buyer who was searching, comparing or asking an AI assistant for a supplier.
The site worked like a brochure. It had strong images but weak headings and unclear categories, so neither a person nor a crawler could tell quickly what was offered. The product pages had little on price, quality, origin or how one stone compares with another, which are the questions a builder, architect or dealer asks before they call. Without that content there was nothing for a search engine to rank and nothing for an AI engine to quote.
The company also had no clear brand entity. Facts about it were inconsistent from place to place, and few authoritative sources mentioned it. Its marketing lived mostly in social posts, which are useful for attention but do not build up into a searchable knowledge base the way structured pages do.
Finally, there was no way to measure the AI side. Ordinary rank reports show where a page sits in Google. They cannot show whether ChatGPT, Gemini, Perplexity or Copilot mention a company. So the company could not tell whether it was visible in the places a growing share of buyers now start.
The build
Each strand supports the others.
Strands one and two
Make it easy to see what is offered and to compare it.
We built the website from the ground up around how people look for stone. That means clear categories, readable headings and a structure that tells both a visitor and a crawler where each product group lives. A buyer should reach the type of stone they want in a few clicks, and a crawler should find the same path in the internal links.
Product architecture is the part many trade sites skip. Each product group gets its own page with a clear name and a consistent set of facts: what the stone is, how it is used, how it differs from alternatives and what affects the price. Pages link to related products and to the guidance a buyer needs. This turns a catalogue of photos into a body of content that can answer questions.
We take care over naming. A trade has its own words, buyers use other words and AI engines may use a third set. A page that uses the buyer words in headings and the trade words in the detail can serve all three.
Strands three and four
Answer the questions buyers ask, on a site engines can read.
The content fills the gaps a brochure site leaves. For each product group we cover price, quality, origin and comparison in plain language, with the practical detail a buyer needs to decide. The aim is pages that stand on their own: someone who lands on one should learn enough to take the next step without searching elsewhere.
Underneath, technical SEO makes sure the pages can be found and understood. That covers crawl access, a clean sitemap, correct canonical tags, sensible internal links, page speed and structured data that matches what the page shows. A fast, indexable site is a precondition. Without it, content effort is spent on pages that never get a fair chance.
We also fix tracking so that visits and enquiries can be attributed. This sounds minor. It is what makes it possible to see which pages produce real conversations rather than just traffic.
Strands five and six
Be the clear, trusted answer, and be mentioned elsewhere.
For generative engines the aim is to make the company easy to identify and easy to cite. We keep facts consistent across the site and wherever else the company is described, and we write answer style sections that give a direct response to common questions before the detail. Google states that AI Overviews and AI Mode need nothing special beyond normal SEO and eligible, indexable pages, so we do not chase hidden tricks.
Authority is the slow strand. Engines and people both trust a company that other credible sources mention. We work steadily on earning mentions and links from relevant sources. We do not buy links and we do not build scaled spam. Our posture on this is firm, and it is described on our About page.
We saw one supporting signal in the citation data: the top cited domain report lists sudarshanstoneix.com first, with 232 citations. We treat that as one signal from one tool, not as proof of the whole outcome.
Screenshots
Measurement
We use two kinds of source and keep them apart. Google Search Console is first party data from Google. For the six month window it reports 17.1K clicks, 1.48 million impressions, a click through rate of 1.2 percent and an average position of 4.9. Its generative AI features report shows 322 thousand impressions over the same period. The top keyword report shows first and third position rankings for commercial granite searches.
The second source is a third party AI visibility tool. It reports 909 mentions across ChatGPT, Gemini, Perplexity, Copilot and Google AI surfaces. An earlier baseline from the same kind of tracking showed zero tracked mentions. That before and after pairing is why we record a baseline at the start of every engagement. It turns a count into a change.
Two cautions apply. Impressions are not visits, and the 1.2 percent click through rate is a reminder that most impressions do not become clicks. The 909 mentions depend on the tool and the prompt set used, and a different tool could report a different number.
Reading the data
1.48M
impressions in six months
How often the site appeared in Google results. A measure of visibility, not of visits.
What moves the price
17.1K
organic clicks
How many times searchers chose the site. Closer to demand, still not an enquiry.
What moves the price
909
tracked AI mentions
Mentions counted by one tool across several engines. Useful as a trend, weak as an absolute.
What moves the price
These are reading notes, not prices.
Limits
It shows a sound foundation, deep product content and steady authority work can create large visibility in a trade with many strong but poorly indexed suppliers.
AI mention counts depend on the tracking tool and prompt set. Treat 909 as a trend marker.
The report covers visibility and clicks. It does not claim a revenue figure, and we do not publish one here.
Website, content, technical work and authority moved in parallel, so we cannot split the credit between them.
Hindsight
The zero baseline is what makes the change visible. We now record it in all five engines before any work begins.
Clicks are not sales. We would set up enquiry tracking per page from day one so revenue questions have an answer.
One tool gives one view. Comparing two would show how much of the 909 is the tool and how much is the market.
What the research says
Checked figures from published studies, each with its source and date. More in our AI search statistics.
37.9% of URLs cited in Google AI Overviews also appeared in the top 10 SERP blocks (37.1% counting only blue links); 31.0% did not rank in the top 100 at all.
Sample: 863K keyword SERPs, ~4M AI Overview URLs. Period: Post-Gemini 3 (from Jan 2026); exact dates not given
Google says optimizing for generative AI search "is optimizing for the search experience, and thus still SEO," with no need to chunk content into tiny pieces or write a specific way for AI.
Sample: Official guidance
Ahrefs found AI Overviews reduce the organic click-through rate for position-one content by 58% (December 2025 vs December 2023), up from the 34.5% reduction it reported in April 2025.
Sample: 300,000 keywords (150k with AI Overview, 150k informational without); aggregated Google Search Console desktop CTR. Period: December 2023 vs December 2025
FAQ
A premium looking website and good products, but weak headings, unclear categories, thin product content, no clear brand entity and no way to measure AI mentions.
The Search Console figures describe Google organic search. They do not include paid clicks. The report does not make a revenue claim, and we do not publish one on this page.
They come from Google Search Console and third party visibility snapshots, and the screenshots on this page are those reports.
No. A mention is a count of times an AI answer named the company in a tracking tool. It depends on the tool and the prompts, and it is not a visit or a sale.
We cannot promise any number. The result reflects a starting point, a trade with a gap to fill and six months of work. An audit shows what is realistic for you.
No. We earn mentions and links from relevant sources, and we do not buy links or build scaled spam. The About page lists what we will not do.
Keep reading
AI assistants lean slightly toward newer pages, but only real updates count. What the studies show, which pages to refresh, and how to keep dates, schema and sitemaps honest.
Read moreAI engines describe your brand from whatever sources they read, and they choose when sources conflict. How to build an entity fact sheet, audit every source and use Organization markup without overrating it.
Read moreAn ordered GEO checklist built from real audits: crawler access, JavaScript rendering, page structure, answer quality, brand consistency, authority and measurement, with how to verify each check and what to fix first.
Read moreYour audit
Message us your website and get a free AI check.
WhatsApp: +91 63777 67206
Email: hello@thegrowingai.com
