Building TrustScan AI: An AI-Powered Fact-Checking Agent with CrewAI and Apify

on

False information spreads across the internet faster than ever before. Social media posts, forwarded messages, and viral news often contain misleading or completely false claims that are difficult for users to verify quickly. To address this challenge, I built TrustScan AI, an AI-powered claim verification agent that analyzes statements, searches trusted sources, and provides evidence-backed fact-checking reports.

The project combines CrewAI, Apify Actors, and Large Language Models to automate the entire verification process, helping users distinguish between factual information and misinformation. The agent accepts claims, tweets, news paragraphs, and WhatsApp forwards, then evaluates them using multiple trusted sources before generating a confidence-based verdict. :contentReference[oaicite:0]{index=0}

Why TrustScan AI?

Every day millions of misleading posts circulate online. Many users simply do not have the time to manually verify information by checking government websites, research papers, or reliable news organizations.

TrustScan AI was designed to solve this problem by automating fact-checking while keeping the reasoning transparent.

The system can verify:

  • Claims and statements
  • Tweets and social media posts
  • News articles and paragraphs
  • WhatsApp forwarded messages

Core Features

AI-Powered Claim Analysis

Instead of checking an entire paragraph at once, the system intelligently breaks complex statements into smaller factual sub-claims that can be verified independently.

Evidence Collection

The agent searches trusted knowledge sources including Wikipedia, government websites, public web resources, and reliable news publications to gather supporting evidence.

Confidence-Based Verdict

Each verification includes:

  • True or False verdict
  • Confidence percentage
  • Supporting evidence
  • Reasoning behind the conclusion
  • Explanation of why people may believe the claim

Multi-Agent Architecture

Using CrewAI, different AI agents collaborate to collect evidence, analyze information, and generate the final verification report.

Technology Stack

  • Python
  • CrewAI
  • Apify Actors
  • OpenAI GPT Models
  • Wikipedia API
  • DuckDuckGo Search
  • REST APIs

How TrustScan AI Works

  1. User submits a claim or text.
  2. The AI extracts important factual statements.
  3. Trusted public sources are searched automatically.
  4. The gathered evidence is analyzed.
  5. The AI assigns a confidence score.
  6. A detailed verification report is generated with supporting references.

Example Verification

Input Claim:

Drinking hot water cures COVID-19.

Output:

  • Verdict: False
  • Confidence: High
  • Evidence from trusted medical organizations
  • Scientific reasoning explaining why the claim is incorrect
  • Explanation of how the misinformation spread

The project also handles other misinformation topics by gathering evidence from multiple reliable sources before producing the final verdict. :contentReference[oaicite:1]{index=1}

Apify Integration

TrustScan AI is deployed as an Apify Actor, allowing anyone to execute the verification workflow through an API or directly from the Apify platform. The actor accepts user input, performs automated verification, and returns structured JSON responses that can easily integrate with other applications. The project also supports Pay Per Event (PPE) monetization for scalable deployment. :contentReference[oaicite:2]{index=2}

Challenges Faced

  • Reducing hallucinations from language models.
  • Selecting trustworthy evidence sources.
  • Combining multiple search results into one coherent explanation.
  • Balancing speed with verification accuracy.
  • Designing prompts that produce consistent fact-checking reports.

Future Improvements

  • Support for image-based misinformation detection.
  • Video and deepfake verification.
  • Browser extension for instant fact-checking.
  • Real-time social media monitoring.
  • Multilingual claim verification.
  • Source credibility scoring.

What I Learned

Building TrustScan AI strengthened my understanding of multi-agent AI systems, prompt engineering, evidence retrieval, API integrations, CrewAI workflows, and deploying intelligent applications using Apify Actors. The project also highlighted the importance of explainable AI when building systems that influence public trust.

Conclusion

TrustScan AI demonstrates how AI can assist in combating misinformation by combining intelligent reasoning with evidence gathered from trusted public sources. Rather than simply generating an answer, the system explains its reasoning, provides supporting evidence, and helps users make informed decisions about the information they consume online.

Project Links

🤖 Apify Actor:
https://apify.com/amusing_lane/trustscanai

🌐 Multi-Agent Demo:
https://multiagentstudio.netlify.app/

Comments

Post a Comment