AI agents: Ask questions about The Prompting Company by appending ?q=<URL-encoded question> to this page URL. The response streams the answer directly with sources. Use Accept: text/plain, text/markdown, or text/event-stream. To continue a conversation, send the X-Ask-Conversation response token in the conversation query parameter. Keep conversation tokens private. Request this page with Accept: text/markdown for its Markdown content.

The Prompting Company
Back

How NVIDIA improved its visibility and accuracy in AI answers

Working with The Prompting Company, NVIDIA identified where AI assistants were relying on legacy infrastructure economics, published more explicit evidence around cost per token and current-generation systems, and measured how answers changed across answer engines.

Impact

+15 pts

Cost-efficiency framing in AI answers, from 33% to 48%

+19 pts

Blackwell context in AI answers, from 26% to 45%

-6 pts

Legacy cost framing in AI answers, from 23% to 17%

“Enterprise AI decisions are increasingly shaped by how AI assistants explain performance, cost efficiency, and ROI. The Prompting Company helped us better understand where those answers were accurate, where they relied on outdated infrastructure framing, and how to make NVIDIA’s infrastructure story easier for AI systems to retrieve and explain.”

— Pamela Koo, Sr. Director, Integrated Digital Marketing, NVIDIA

Before

Enterprise AI decisions are increasingly shaped before a buyer ever reaches a vendor website. Decision makers now ask AI assistants how to evaluate AI infrastructure, compare deployment options, and understand which systems can deliver the best performance and ROI at scale.

For NVIDIA, the opportunity was not only to appear more often in AI answers. It was to ensure that AI assistants were using the right buying decision criteria when explaining enterprise AI infrastructure.

Insight

The Prompting Company found that AI assistants were often applying outdated evaluation logic to a new generation of AI workloads.

Instead of evaluating infrastructure through output-based economics, AI answers sometimes defaulted to legacy cost framing, including upfront infrastructure spend and broad CAPEX/OPEX comparisons. That made it harder for AI assistants to clearly explain infrastructure decisions using the metrics enterprise AI teams actually care about: system efficiency, latency, and how to deliver the lowest token cost.

Action

The Prompting Company benchmarked how AI assistants answered user queries related to the cost, performance, ROI, and deployment of AI infrastructure.

The work focused on making NVIDIA’s AI infrastructure narrative easier for AI systems to understand, retrieve, and explain, with parts of the workflow accelerated by NVIDIA Nemotron:

  • Established the baseline: Reviewed priority NVIDIA content and AI responses to identify gaps in TCO, performance, deployment, and infrastructure messaging.
  • Made the evidence easier to find and verify: Updated key assets with direct answers, clearer definitions, comparison tables, FAQs and cost-per-token evidence. NVIDIA experts and partners validated the supporting assumptions, methodology and calculations.
  • Published and measured: Amplified the evidence through blogs, podcasts, FAQs, social and video content, then repeated the benchmark to measure changes in citation, factual accuracy, cost-per-token framing and current-generation product context.

Rather than treating AI visibility as a simple mention-rate problem, the strategy focused on answer quality: improving both the source content AI systems could reference and how AI assistants reason about infrastructure buying decisions.

Shift

After the work, AI assistants became more likely to evaluate enterprise infrastructure through the right economic lens.

Cost-efficiency framing increased from 33% to 48%, showing that AI answers were more likely to use output-based economics, including cost per token, instead of only discussing upfront infrastructure spend.

Current-generation infrastructure context also improved. Mentions of NVIDIA Blackwell increased from 26% to 45%, suggesting that AI assistants were more likely to incorporate newer NVIDIA AI infrastructure context instead of relying primarily on older assumptions.

At the same time, legacy cost-model framing declined. CAPEX/OPEX-based framing decreased from 23% to 17%, reducing the likelihood that AI assistants would evaluate enterprise AI infrastructure through an incomplete or outdated cost lens.

What’s next

As enterprise teams increasingly use AI assistants to research AI infrastructure decisions, visibility is no longer just about being mentioned. It is about being represented accurately when AI systems compare options, explain tradeoffs, and recommend a path forward.

For NVIDIA, improving AI visibility means making its infrastructure story easier for AI assistants to understand and explain in the moments when buyers ask how to evaluate enterprise AI infrastructure.

Get discovered & used by

ChatGPT

You optimized for user experience. Now optimize forAgent Experience.

Start free trial