AI Visibility Audit Tool

The problem. Existing AI visibility tools only report whether a business shows up when people ask ChatGPT or Claude, they don't fix the underlying reasons it doesn't.

What I built. A CLI tool, solo end to end: a six-stage pipeline that generates realistic buyer-intent questions a prospective customer might ask an AI assistant, actually queries ChatGPT, Claude, Perplexity, and Gemini with live web search enabled, extracts whether and how the business gets mentioned, separately audits the business's own site for the technical reasons AI crawlers might be missing it, and outputs both a plain-English report and ready-to-paste fixes (structured data, FAQ blocks, rewritten page copy) rather than stopping at a diagnosis.

A decision worth noting. Rebuilt all four provider integrations to use live, vendor-native web search after verifying that ungrounded API calls understated real visibility: a direct consumer-app query surfaced a business that the bare API call missed entirely. Deliberately excluded prompts containing the business's own name from the headline visibility score, since a model answering a branded question is mostly echoing a name it was handed rather than surfacing it organically; on real test data the blended rate read around 30% while the organic-only rate was 3–8%, a difference that changes the actual takeaway a client walks away with. Also dropped competitor score comparisons entirely after the same competitor pair's score swapped from 25 to 4 depending only on which business was the audit subject: recognized that a flat score next to a named competitor reads as a factual claim a small prompt sample can't support, and chose defensibility over a punchier chart.