Skip to main content

Transform Claims with AI-Powered Fact-Checking

CheckThat AI converts noisy social media posts into precise, verifiable claims using advanced language models and iterative refinement with G-Eval feedback.

Get StartedAPI Reference
8+ AI Models
G-Eval Quality Assessment
Real-time WebSocket
# Input claim
”The government is hiding alien technology!”
# Normalized output
”Government transparency concerns raised regarding alleged undisclosed technology programs.”
# Quality score
METEOR: 0.847 | G-Eval: 0.89

Quick start

Get CheckThat AI up and running in minutes

1

Install dependencies

Clone the repository and install required packages for both frontend and backend.
git clone https://github.com/Nikhil-Kadapala/checkthat.git
cd checkthat
./setup-project.sh
2

Configure API keys

Set up your AI model API keys as environment variables.
export OPENAI_API_KEY="your-openai-key"
export ANTHROPIC_API_KEY="your-anthropic-key"
export GEMINI_API_KEY="your-gemini-key"
You only need API keys for the models you plan to use. Llama 3.3 via Together.ai works without authentication.
3

Start the application

Launch both the FastAPI backend and React frontend.
./run-project.sh
The web interface will be available at http://localhost:5173 and the API at http://localhost:8000.
4

Normalize your first claim

Use the chat interface or make a direct API call to normalize a claim.
curl -X POST "http://localhost:8000/chat" \
  -H "Content-Type: application/json" \
  -d '{
    "user_query": "Climate change is a hoax invented by scientists",
    "model": "meta-llama/Llama-3.3-70B-Instruct-Turbo-Free"
  }'

{
  "normalized_claim": "Scientific consensus on climate change contested by skeptics",
  "confidence_score": 0.85,
  "model_used": "meta-llama/Llama-3.3-70B-Instruct-Turbo-Free"
}

Explore by topic

Learn about CheckThat AI’s capabilities and features

Claim normalization

Understand how noisy social media posts are transformed into fact-checkable statements

Refinement pipeline

Learn about iterative claim improvement with Self-Refine and Cross-Refine algorithms

Supported models

Explore the 8+ AI models available including GPT-4, Claude, Gemini, Llama, and Grok

G-Eval metrics

Deep dive into quality assessment using DeepEval’s G-Eval framework

Features and capabilities

Everything you need for claim normalization and fact-checking

Web interface

Interactive chat interface for real-time claim normalization

Batch processing

Process multiple claims simultaneously with progress tracking

API access

RESTful API with WebSocket support for streaming responses

Multiple strategies

Zero-shot, few-shot, and chain-of-thought prompting techniques

Quality metrics

METEOR scoring and G-Eval assessment for claim quality

CLEF CheckThat

Built for CLEF-CheckThat! Lab Task 2 competition

Ready to start fact-checking?

Get started with CheckThat AI today and transform how you verify claims from social media.

View quickstart guide

Build docs developers (and LLMs) love

Get started for freeTalk to us