Last Updated on September 25, 2026 by Amit Kakkar
A cloud GPU provider was competing for AI and ML teams in one of the fastest-moving categories in tech. Its buyers, engineers and technical founders, increasingly ask ChatGPT, Perplexity and Gemini which GPU cloud to use, how pricing compares, and which provider suits a given model or workload. The brand needed to be part of those answers, not just the Google results underneath them.
| Metric | Before | After | Change |
|---|---|---|---|
| AI brand mentions | 30 | 120 | 4x |
| Pages cited by AI platforms | 200 | 500 | 2.5x |
| AI visibility score | 10 | 17 | +70% |
| Monthly organic search visitors | 700 | 9,000 | +1,186% |
How we measured it : AI citations were tracked in Semrush. For prompt tracking, we ran a fixed set of buyer prompts through the DataForSEO API to monitor how often, and where, AI assistants named the brand.
The challenge : a technical product AI could not read well
GPU cloud is a crowded category, and AI assistants tend to default to the biggest hyperscalers unless they have clear, trustworthy reasons to recommend someone else. When we started tracking, the client had just 30 AI brand mentions and a visibility score of 10, so it was named only occasionally when buyers asked AI for recommendations. Four problems were holding it back.
- Crawl and indexing issues : Slow pages, crawl errors and improper indexing meant search engines and AI crawlers could not reliably reach or read key pages. If a page is not crawled, it cannot be cited.
- Confusing site structure : Weak internal linking and unclear navigation left product, pricing and GPU pages disconnected from the educational content that explains them. AI systems had no clear map of what the company offered or how its pages related.
- No content built around buyer questions : There was no keyword or prompt strategy, so content did not answer the specific questions engineers ask, like which GPU fits a workload, what it costs per hour, or how providers compare.
- Weak third-party trust signals : The backlink profile was full of spammy, irrelevant links. AI models lean on credible outside sources to decide which brands to recommend, and this profile gave them little to work with.
Our GEO strategy
The goal was to make the client the obvious answer when an engineer asks an AI assistant about GPU cloud. We worked on four fronts.
1. Making the site machine-readable
We added schema markup to product and GPU pages so AI systems could clearly understand what each offering is, and published an llms.txt file to point AI crawlers to the most useful pages. Combined with the crawl and indexing fixes, this gave AI platforms a clean, structured view of the product.
2. Answering the questions buyers ask AI
We mapped the questions AI and ML teams ask when choosing a GPU cloud, then built and reworked content to answer them directly. That included clear, self-contained answers under descriptive headings, FAQ sections on key pages, and technical tutorials that explain how to run real workloads. Each page was written so any section could be lifted into an AI answer and still make sense.
3. Building topical authority with a connected structure
We rebuilt internal linking so product pages, GPU pages, tutorials and blog content support each other. Content gap analysis showed where competitors were covering topics the client was not, and we filled those gaps. We also resolved keyword cannibalization so each page had one clear job instead of several pages competing for the same question.
4. Earning credible third-party signals
We cleaned up the spammy backlink profile and replaced it with links from high-authority industry sites (DA 80+) through outreach and linkable assets. We also studied how the category is discussed on Reddit, where engineers compare providers openly and which AI models draw on heavily, to shape positioning and content around what buyers actually care about.
The SEO foundation behind it
None of the GEO work lands on a site crawlers cannot read, so the technical fixes came first: a full audit covering page speed, crawl errors, robots.txt and sitemaps, then on-page work on meta tags, headers, internal links and mobile experience. Keyword research focused on high-potential, lower-competition terms across short and long-tail queries.
The same work that made the site easier for AI to understand also lifted it in Google. Monthly organic traffic grew from 700 to 9,000 visitors within a year, a 1,186% increase.
Results
- 120 AI brand mentions, up from 30 (4x) : The brand now appears by name far more often when AI assistants answer GPU cloud questions, alongside much larger competitors.
- 500 pages cited by AI platforms, up from 200 (2.5x) : Tutorials, GPU pages and educational content are being used as sources in AI-generated answers.
- AI visibility score of 17, up from 10 (+70%) : Measured across our tracked set of buyer prompts.
- 1,186% organic traffic growth : Monthly organic visitors grew from 700 to 9,000, backed by DA 80+ links and a cleaner backlink profile.
What this means for AI infrastructure companies
Technical buyers now research inside AI assistants before they ever compare pricing pages. In a category dominated by hyperscalers, the providers that get named in those answers are the ones that make their product easy for AI to understand, answer real engineering questions clearly, and earn trust from credible outside sources. That is repeatable, and it is what this project was built on.
See where your brand stands in AI answers. Get a free AI Visibility Plan we test 20 buyer prompts across ChatGPT, Claude, Gemini and Perplexity and show you your visibility score and citation gaps.
About the Author
