FactSnap – AI Powered Fact-Checking App

FactSnap is your AI-powered Fact Checker—instantly verifying any claim using real-time web evidence and fast LLMs + RAG system.

Comment

Inspiration

In a world where misinformation spreads faster than facts, we wanted to build something that helps people verify claims instantly—with trust, speed, and context. Whether it's a trending tweet, a viral video, or a WhatsApp forward, FactSnap was created to bring clarity and confidence to everyday information.

What it does

FactSnap is an AI-powered fact-checking web app that lets users input any claim and instantly get:

  • A verdict (True / False / Partially True / Unverifiable)
  • A confidence level (High / Medium / Low)
  • A clear explanation with reasoning
  • 🌐 Web evidence fetched in real time via Brave Search API (RAG system)
  • 📚 Chat history, 🔁 follow-up support, 📤 screenshot sharing, and 🗣️ voice input

It’s like having a personal fact-checking assistant—powered by top LLMs and live internet data.

How I built it

We used the Bolt.new AI-first platform to build and deploy the entire application. The key technologies and architecture include:

  • Frontend: React + TypeScript, Tailwind CSS
  • AI Backend: Groq API (LLaMA 3, Mistral, Gemma)
  • RAG System: Brave Search API integrated to fetch real-time web context based on query intent
  • Authentication & DB: Firebase (Google Sign-In + Firestore)
  • Voice Input: Web Speech API
  • Screenshot Sharing: HTML2Canvas
  • Hosting: Netlify We implemented intelligent routing that detects when a query needs recent info and dynamically triggers web search to enhance AI output.

Challenges I ran into

  • 🔄 Managing dynamic fallback between normal Groq output vs RAG-enhanced output
  • ⚙️ Debugging issues with the Brave API (429 errors, timing glitches)
  • 🧠 Prompt engineering for Groq to determine query intent
  • 💾 Making chat history work seamlessly across Brave vs non-Brave responses
  • 💬 Ensuring follow-up questions preserve context correctly—both when web search is used and when it's not.
  • 🌐 Debugging inconsistencies between Bolt preview environment and deployed version, especially around Brave API responses.
  • 🎬 Learning and editing the demo video using iMovie as a complete beginner, handling tasks like voice syncing, trimming, overlays, and export formatting.

Accomplishments that I am proud of

  • Successfully implemented a fully working RAG system using Brave Search API
  • Built a polished, responsive, and performant product in a short time using Bolt
  • Achieved fast, relevant fact-checks that feel like magic to users
  • Integrated voice input, follow-ups, and screenshot sharing for real UX impact
  • Created a professional-grade 3-minute video demo to showcase the project

What I learned

  • How to leverage LLMs and RAG together in a real-time app
  • Effective prompt engineering to improve output structure and control behavior
  • How to build fast in Bolt.new and adapt to real-world API constraints
  • That user experience and clarity are just as important as core functionality

What's next for FactSnap – AI Powered Fact-Checking App

  • 🧠 Add multimodal input (image-based fact-checks)
  • 🗂️ Improve history search and tagging
  • 🌍 Support more regional languages
  • 🔒 Add trust scoring for sources
  • 🚀 Launch it publicly and help fight misinformation at scale

Built With

Share this project:

Updates