We send twenty-five real questions to five live AI assistants — the same questions a homeowner would type — and grade what comes back. No scraping, no model training, no opinions. Four steps, every one of them yours to verify.
Step 01 / 04
A single fetch of your homepage. We check fifteen on-page signals — schema markup, meta tags, content depth, technical hygiene — and compute a preliminary score. This is the baseline.
No crawler. No login. A polite, single GET request.
Step 02 / 04
Five prompts per AI, drawn from four categories that map to actual customer behavior: generic local ("best plumbers in Tampa"), branded ("tell me about Joe's Plumbing"), problem-led ("who fixes burst pipes in Tampa"), and comparative ("top 5 plumbers in Tampa").
Claude · ChatGPT · Gemini · Perplexity · Google AI Overviews
Step 03 / 04
Each response is scanned for mentions of your business name, domain, or canonical variants. Each AI is graded separately on a familiar A-to-F scale, then combined into one overall index.
Fuzzy matching handles spacing, casing, and common abbreviations.
Step 04 / 04
The results page shows every query, every AI's full verbatim response, every competitor that got recommended instead of you, and every on-page signal we found missing. No black box, no editorial spin.
Sharable by link. Stored as static JSON in Vercel Blob.