DecetAI

DecetAI analyze any online text — from tweets to news articles — and instantly cross-verifies it with credible sources like Wikipedia, The Guardian, and academic papers.

Comment

Inspiration

Every day, we’re flooded with online content, including news, social posts, and articles. It’s hard to tell what’s true. I wanted a tool that could instantly fact-check anything I read, without having to open ten tabs or verify sources myself. That frustration inspired DecetAI, an AI-powered misinformation checker that helps people separate fact from fiction in real time.

What it does

DecetAI is a web application, and soon to be Chrome Extension, that allows you to analyze its truthfulness. It...

  • Extracts keywords and context from the highlighted text
  • Cross-verifies claims using credible sources like Wikipedia, The Guardian, arXiv, and real-time news APIs.
  • Uses Azure OpenAI to evaluate the accuracy of the statement based on evidence from these sources.
  • Returns a clear, summarized explanation right inside the popup — no need to leave the page.

How we built it

  • Frontend: Modern and lightweight UI built with HTML, CSS, Tailwind CSS, and JavaScript
  • Backend: Python Flask API running locally, handling text analysis and AI processing.
  • AI & NLP: Used KeyBERT for keyword extraction. spaCy for text processing and sentence relevance detection. Azure OpenAI GPT model for misinformation reasoning.
  • APIs integrated: Google Custom Search (for credible web sources) NewsAPI (for recent news) The Guardian API, arXiv API and Wikipedia API (for factual context)

Challenges we ran into

  • Integration of CORS and browser extension permissions with a local Flask server.
  • Some sites either rejected requests or necessitated sophisticated parsing, thus making the extraction of trustworthy data from news and research APIs a tedious process.
  • Maintenance of rate limits and latency when joining together different APIs and an LLM.

Accomplishments that we're proud of

  • A system that can actually fetch and assess real proof — not only viewpoints — was established.
  • Achieved end-to-end integration between Flask, OpenAI, and browser APIs in the little time that was allotted for the hackathon!

What we learned

  • An effective way to integrate frontend browser extensions with AI-powered backend systems.
  • A better understanding of natural language processing for the purpose of misinformation detection.
  • A lot of knowledge in API orchestration and how to make noisy search results structured and analyzable data.
  • When combating misinformation, designing for trust and clarity is equally important as designing for accuracy.

What's next for AI-Misinformation-Checker

  • Complete and design a modern interface for the Chrome Extension that would allow you get instant feedback on truthiness while reading articles, news, blogs, and more.

Built With

  • arxivapi
  • azureopenai
  • chromeextensions
  • css
  • flask
  • googlecustomsearchapi
  • html
  • javascript
  • keybert
  • mediawikiapi
  • newsapi
  • python
  • spacy
  • tailwindcss
  • theguardianapi
Share this project:

Updates