Show HN: Open-source tool for reverse engineering ChatGPT queries about brands
An open-source tool has been released that estimates prompts users might give to ChatGPT and other AI platforms based on a website's content. It crawls sites, analyzes keywords, and finds similar conversations in the WildChat dataset to reverse-engineer how brands are referenced in AI responses. The tool is aimed primarily at e-commerce stores looking to understand AI visibility.
Background
- **GEO / AI SEO** (Generative Engine Optimization) is an emerging practice of optimizing content so that AI chatbots like ChatGPT, Gemini, or Perplexity cite or recommend a brand in their answers — distinct from traditional search-engine SEO.
- **WildChat** is a publicly released dataset of ~1 million real, anonymized ChatGPT conversations, collected from users who interacted with a free ChatGPT wrapper. It's frequently used in research to study real user behavior.
- **No native analytics**: Chat platforms currently don't provide brand owners with any dashboard showing how often or in what context the AI mentions their products, making it hard to measure AI visibility.
- **Reverse-engineering prompts**: Since the platforms are opaque, this tool tries to infer what questions users are actually typing that lead an AI to mention (or fail to mention) a given brand, by matching a website's topics against real conversations in WildChat.