LLM Brand Audits That Keep the Receipts

What it does

NiubiGEO is a self-hosted TypeScript application that audits how visible your brand is inside LLM answers. You enter a domain, choose models through OpenRouter, and run tests to collect each model’s description of your product, the competing products it names, and the sources it cites. You can also run keyword tests—asking about “prototyping” without naming Figma, for example—to see which brands appear unprompted.

The interesting bit

Rather than distilling AI visibility into a proprietary score, NiubiGEO preserves the full forensic record: original answers, citations, failures, and test conditions are all kept on record so you can inspect the evidence yourself. The repository includes 20 real-world cases for products like PostHog, Notion, and Figma, letting you explore actual model outputs before you even install the tool.

Key highlights

• Open source under Apache-2.0; runs self-hosted via Docker or Node.js 22+
• Query multiple models via OpenRouter with independent web-search toggles per model
• Domain tests surface brand descriptions and competitor mentions; keyword tests reveal who gets recommended when your brand is not named
• Stores original responses, provider citations, and failure logs for every run
• 20 documented real cases with screenshots, test conditions, and unresolved findings
• Optional scheduled monitoring worker to collect repeated measurements over time

Caveats

• You supply your own OpenRouter API key and cover model inference and search costs
• The authors explicitly warn that a few minutes of repeated tests do not establish long-term growth trends
• At v0.2.0, the project is early-stage and openly records failures and unresolved findings alongside successes

Verdict

Product and marketing teams who want to investigate AI visibility with primary evidence rather than trust a black-box score. If you need a fully managed SaaS with built-in marketing workflows, the repo maintains a comparison table of commercial alternatives for that.