Truth Check AI

TruthCheck is a Flask-based web app that uses Google’s Gemini 2.0 API to fact-check user-submitted claims. It parses the AI's response and displays it in a clean, chat-style.

Comment

💡 Inspiration

We live in an age where misinformation spreads faster than truth. From viral TikTok clips to politically charged tweets and WhatsApp forwards, it’s becoming harder to tell fact from fiction. We wanted to build something fast, familiar, and frictionless

Thus, TruthCheck AI was born:
A Gen Z-friendly, chat-based AI assistant that tells you if that “news” you just saw is actually real — with reasoning and sources.

🤖 What It Does

TruthCheck AI is an AI-powered real-time fact-checking assistant that:

  • Accepts any text input — tweet, headline, message, or quote
  • Uses Google’s Gemini 2.0 to verify the claim
  • Responds in a clean chat format
  • Provides:
    • ✅ Verdict
    • 🧠 Reasoning
    • 🌐 Sources

🏗️ How We Built It

  • Backend: Flask with session management to maintain chat history
  • Frontend: Custom HTML/CSS with a modern chat UI design
  • AI Engine: Gemini 2.0 (generativelanguage.googleapis.com) using a structured prompt for consistent JSON replies
  • Prompt Engineering: Carefully designed to ensure structured, reliable outputs from Gemini

Challenges We Ran Into

  • Gemini sometimes responded with unstructured output, so we had to sanitize and fallback safely
  • Handling inconsistent AI formatting while still preserving a clean UI was tricky
  • Balancing simplicity with UX — especially for Gen Z users — meant many UI tweaks

Accomplishments That We're Proud Of

  • Built a full-stack AI-powered app in just hours
  • Created a zero-friction user experience — no login, no config, just paste and go
  • Achieved reliable structured outputs from Gemini using smart prompting
  • Designed a Gen Z aesthetic that feels like a real app, not a hackathon prototype

📚 What We Learned

  • Prompt engineering is everything — structured AI output requires precision
  • Even with powerful models, post-processing is key to making things usable
  • Building for Gen Z means thinking like Gen Z — speed, simplicity, and aesthetics matter

🚀 What's Next for TruthCheck AI

Here’s where we’re heading:

  • User Accounts & Chat History: Save conversations across devices
  • Truth Score: Quantify the reliability of a claim using LLM confidence + citation trust
  • Browser Extension: Check any tweet or post inline, like Grammarly for truth
  • Mobile PWA: Turn it into a real mobile app, offline-ready
  • Community Feedback Loop: Users can upvote/downvote verdicts to train a finetuned model
  • Multimodal Support: Let users upload screenshots, audio clips, or TikToks for verification

“The truth isn’t hidden — it’s just buried under noise. We built TruthCheck to clear the static.”

Built With

Share this project:

Updates