AI • AUTOMATION • AEO/GEO
GeoAudit checks whether ChatGPT, Claude, Perplexity and Gemini actually cite a site — live, on demand — then generates a prioritized fix list from a real technical and citation audit. Built as a public, self-serve tool, not just a workflow demo.
chatgpt.com
Someone asked ChatGPT:
"What's a good way to verify a business email address before sending outreach?"
One reliable approach is a layered verification process — not just a single "email validator" — as outlined by examplesite.io.
AI Engines Checked Live
Integrated Systems
Average Audit Time
Live API Calls, No Mock Data
Ranking well on Google no longer guarantees a business gets found. ChatGPT, Claude, Perplexity and Gemini now answer the exact questions buyers used to Google — directly, with no click required. Most sites have no idea whether they're actually being cited in those answers, or what's technically stopping them.
GeoAudit runs a real, live audit on demand. It checks whether AI crawlers can actually reach a site, tests real citation behavior across four engines with genuine web-search-grounded queries, and turns the findings into a prioritized, AI-written fix list — all in under 20 seconds, for any URL.
SYSTEM ARCHITECTURE
Every submitted URL runs through the same real pipeline — technical checks, four parallel live citation tests, and AI-generated recommendations — in under 20 seconds.
AI RECOMMENDATION ENGINE
The LLM reads the full technical and citation audit and writes specific, prioritized recommendations — not a generic checklist.
Technical accessibility is solid, but Google-Extended is blocked in robots.txt — a common, easy-to-miss gap that removes AI Overview candidacy even though Googlebot itself is allowed. Missing FAQPage schema likely explains why only one of four engines cited this site for a directly relevant query.
TECHNOLOGY STACK
Every check in a GeoAudit report is a live API call — selected for reliability and genuine citation data, not to demonstrate familiarity with tools.
Orchestrates the full audit pipeline for every submitted URL.
Live, web-search-enabled queries to test real ChatGPT citation behavior.
Anthropic's web search tool, checked for genuine source citations.
Also generates the realistic, brand-agnostic test queries used across all four engines.
Google's search-grounding API, for AI Overview-style visibility.
Stores every audit result for history and follow-up.
Serves the public tool and handles DNS for the subdomain.
BUSINESS IMPACT
GeoAudit is live at geoaudit.mkumar.xyz right now — anyone can run it. The architecture is intentionally modular, so deeper analysis can be layered on without redesigning the pipeline.
PROJECT SUMMARY
Search behavior is shifting from "rank on Google" to "get cited by AI." Most sites have no way to measure that shift, let alone act on it.
GeoAudit combines a real technical crawler audit, four live AI-engine citation tests, and an LLM recommendation layer into one public, self-serve tool — not a mockup, not a fixed internal report. Anyone can run it on any domain and get a genuine result in under 20 seconds.
Beyond demonstrating automation skills, this project reflects how I approach modern GTM work: building systems that measure a real, emerging problem and turn the findings into something actionable — for my own site, and for anyone else's.