Check out a quick demo of our prototype:
[Watch the Demo]

• Debunkd utilizes Optical Character Recognition (OCR) powered by Tesseract to extract text from uploaded images.
• This allows even visual content to be analyzed for potential misinformation.

• The extracted text is processed by a Retrieval-Augmented Generation (RAG) model.
• The model converts the text into embeddings and performs a similarity search using a vector database (ChromaDB) to cross-reference facts.

• Debunkd integrates with Google’s Fact Check Tools API to fetch real-world fact-checking data.
• This ensures the results remain accurate and up-to-date.

• The ChatGroq LLM generates human-like responses, explaining the validity of claims and providing additional context.
• Users can interact with the AI-powered chatbot for further verification.

• Users are shown an image and must decide if it’s real or fake.
• A score tracker records user performance.
• High scores are stored and displayed for competitive play.

Users can verify an image in two ways:

1. Upload an image (or select a sample) – The system performs LLM-based analysis & OCR (Tesseract) to assess authenticity.
2. Chatbot for deeper insights – Users can chat with an AI assistant to get explanations and verify claims.

• The chatbot remembers previous messages, making conversations cohesive and intelligent.
• Users can ask follow-up questions and receive context-aware responses.

💬 User: "Joe Biden is dead."
🤖 System: "No, Joe Biden is alive. Here are some reliable sources: …"
💬 User: "Then how old is he?"
🤖 System: "Joe Biden is 81 years old." (instead of generating a generic response)