RateMyStay — AI Hotel Review Analyser
A free AI-powered web tool that analyses hundreds of hotel reviews instantly, delivering a clear score, pros, cons, and a booking verdict in under 15 seconds.
What needed solving
Travellers booking hotels face decision fatigue from reading hundreds of contradictory guest reviews across multiple platforms. There was no free, instant tool that could cut through the noise and deliver a clear, unbiased verdict before booking.
How I solved it
Built a full-stack AI web tool where users search any hotel from a curated global dataset, triggering a Groq-powered Llama 3.3 analysis of real guest reviews. The system returns a structured verdict — score, pros, cons, hidden red flags, and a booking recommendation — in under 15 seconds. A Redis-first caching architecture ensures repeat searches return in under 100ms with zero AI API calls.
A look at the product
Key Features
- AI analyses hundreds of real guest reviews and returns a structured 10-point score, top pros, top cons, and a clear Worth Booking / Maybe / Avoid verdict
- Redis cache layer stores analysis results by hotel ID for 24 hours — 100 users searching the same hotel triggers only 1 AI call
- Multi-layer bot protection combining Cloudflare Turnstile, browser fingerprinting, composite IP+fingerprint rate limiting, honeypot fields, and User-Agent blocking
- 3 free searches per day per user enforced via composite Redis keys — no login or registration required
- Animated floating review cards hero section with SVG grid background communicating the product's core value instantly
- Fully responsive across mobile, tablet, and desktop with a premium light theme and Playfair Display typography
Architecture
- 01Cache-first request flow — Redis checked before every AI call, reducing Groq API usage by up to 99% on popular hotels
- 02Composite identity system using FingerprintJS visitor ID combined with IP address as a single Redis quota key — catches VPN abuse and browser switching independently
- 03Global daily cap of 300 analyses per day stored in Redis with automatic midnight reset to prevent abuse at scale
- 04Structured JSON prompt engineering — single Groq API call returns the complete analysis object including score, verdict, category scores, sentiment breakdown, and quotes
- 05Serverless API routes on Vercel with edge-compatible Upstash Redis client — no persistent server, infinitely scalable
- 06Cloudflare Turnstile invisible widget integrated client-side with server-side secret verification on every analysis request
Results & Outcomes
Sub-100ms response time on cached hotel searches via Redis — effectively instant for repeat queries
99% reduction in AI API calls on popular hotels through Place ID based result caching
Full bot protection across 8 distinct attack vectors with zero friction for real users
3-search daily quota enforced without any user registration, login, or cookies
Successfully deployed and live on Vercel with production Cloudflare Turnstile, Upstash Redis, and Groq AI integration
Want a similar solution?
I'm available for new projects. Let's talk.