- AI
- Frontend
- Automation
- Backend
- Fullstack
klarava – AI understandability check for websites
Image credit: Majd Edriss
Challenge
More and more people ask ChatGPT & co. instead of searching Google. But most websites are hard for AI to read: if you don’t clearly state who you are, what you offer and for whom, you show up wrongly in AI answers, or not at all.
Result
A free tool that shows in seconds how well an AI understands a website (score, gaps and “how AI currently sees you”) and, on request, delivers ready-made building blocks to improve it, fully automated by email. The same engine has also grown into a paid API that returns the identical analysis as machine-readable JSON, sold via RapidAPI.
Solution
klarava crawls the key pages, scores six dimensions via LLM and checks verifiable technical signals (Schema.org, hreflang, llms.txt). The report with ready-made building blocks (AI profile, FAQ, llms.txt, action briefing) is generated and sent automatically, dependency-free in Node, EU hosting, GDPR-compliant. The same engine became a paid API (three score layers, two deterministic) on a dedicated server, distributed via RapidAPI.
Technologies
- Node.js
- OpenAI gpt-4o-mini
- Supabase (EU)
- Vanilla JS
- REST API (OpenAPI)
- RapidAPI
- VPS (systemd + Caddy)
- nodemailer
- Cloudflare Turnstile