Interview

The Prompting Company helps products get mentioned in ChatGPT using shadow sites — lands Nvidia as customer, hits $300K

Sep 10, 2025 with Kevin Chandra

Key Points

  • The Prompting Company builds markdown-based shadow sites that compress web content into a format LLMs can ingest more efficiently than raw HTML, positioning itself as infrastructure rather than content marketing.
  • Nvidia is a paying customer, won through competitive bidding, as the startup reaches $300K in revenue without pivoting from its original YC-batch thesis.
  • An oversubscribed fundraise is underway with a lead investor in selection, though the durability of the shadow site model depends on how AI providers tighten crawl policies and weight training sources.
The Prompting Company helps products get mentioned in ChatGPT using shadow sites — lands Nvidia as customer, hits $300K

Summary

The Prompting Company is a YC-backed startup that helps brands appear in ChatGPT responses by building markdown-based shadow sites mirroring a client's web presence. These sites are formatted for foundation models to ingest far more efficiently than raw HTML. Markdown strips out CSS and formatting overhead, making it cheaper and faster for models to crawl and index the content.

Kevin Chandra, the co-founder, positions this as infrastructure rather than content marketing. The company entered YC with the idea intact and has not pivoted. Traction sits at $300K in revenue, with Nvidia as a paying customer won through a competitive bake-off. The raise is oversubscribed and they are selecting a lead investor.

The shadow site mechanic is their differentiating layer. Rather than generating articles, the company creates parallel model-readable versions of a client's site designed to influence what LLMs surface in answers. This is generative engine optimization built at the infrastructure level rather than the content level. The durability of this approach depends on whether model providers tighten crawl policies or adjust how they weight training and retrieval sources.