Team 45 - Are U ReaL?

• 

  Homepage of Are U ReaL?

• 

  Analysis from Are U ReaL?

Inspiration

Misinformation is everywhere—fake news websites, AI-generated content, and misleading headlines are becoming increasingly common. Many users struggle to determine whether an article or website is trustworthy, often falling for deceptive narratives. Are U ReaL? was built to tackle this problem, helping users verify the authenticity of news sources by analyzing website credibility, cross-checking facts, and detecting manipulated images.

What it does

Are U ReaL? is an AI-powered tool that scans URLs and determines whether a website is spreading misinformation. Here’s how it works: 1️⃣ Users enter a suspicious website link 2️⃣ Our Web Parser Agent extracts text and images from the site 3️⃣ The Image Forgery Expert checks for manipulated visuals 4️⃣ The News Analyst Agent analyzes headlines and article structures 5️⃣ The Web Researcher Agent cross-checks sources via DuckDuckGo 6️⃣ The Misinformation Expert flags potential fake news patterns 7️⃣ If necessary, we reference a blacklist of known fake news sites 8️⃣ The results are summarized and presented to the user

How we built it

We used CrewAI to manage multiple agents working together in an automated pipeline. Our system consists of:

Agents 🔹 Web Parser – Extracts text and images from the target website 🔹 Image Forgery Expert – Detects manipulated or AI-generated images 🔹 News Analyst – Evaluates headlines and body content for misleading patterns 🔹 Web Researcher – Cross-checks claims using search engines 🔹 Misinformation Expert – Flags potential fake news patterns 🔹 Blacklist Checker – Compares URLs against a known list of misinformation sites

Technologies Used 🛠 CrewAI – A framework for building AI-powered agent teams that can complete complex tasks 🛠 OpenAI – LLM-powered text analysis and misinformation detection 🛠 NeonDB – Maintains a credibility index of known sources 🛠 Tavily – Enables efficient web searching for fact-checking 🛠 Firecrawl – Scrapes and extracts content from input URLs 🛠 Render – Hosts and deploys our web-based verification tool

Challenges we ran into

Credibility validation: Even legitimate news sources can sometimes publish misleading information, making it hard to assess trustworthiness.

Efficient web search: Cross-checking facts without excessive API costs or unnecessary noise.

False positives: Some articles from credible sources can still appear misleading due to their wording.

Ensuring transparency: Making sure users understand why a site was flagged instead of giving a "black box" judgment.

Accomplishments that we're proud of

✅ Successfully integrated multiple AI agents into a cohesive misinformation detection system ✅ Created a user-friendly website scanner that simplifies fact-checking for everyday users ✅ Implemented image forgery detection to identify manipulated visuals in fake news articles

What we learned

How to orchestrate multiple AI agents efficiently with CrewAI. The complexity of misinformation detection, especially in ambiguous cases. The importance of fact-checking against multiple sources instead of relying on a single authority. How to reduce bias in LLM-based assessments through step-back prompting

What's next for Are U ReaL?

Built With

• crewai
• fastapi
• firecrawl
• neondb
• next.js
• openai
• python
• tavily