Trust Issues

Trust Issues in Action
Start Page
Searching...
Summary Result

Inspiration

In today’s digital environment, misinformation spreads faster than verification. With the rise of generative AI, it has become increasingly difficult to distinguish between authentic reporting, manipulated content, and fully synthetic media. We were inspired by the idea of creating a digital investigation tool a tool something that treats online content like a case to be examined rather than information to be passively consumed.

We also wanted to address a practical hackathon constraint: AI-powered verification tools can quickly become expensive due to API usage. That pushed us to think about efficient verification workflows and caching strategies.

What it does

Trust Issues is a browser extension that analyzes the content currently visible on a webpage and generates a credibility report.

When the user clicks Scan This Page, the extension:

Extracts visible page text
Identifies core claims
Verifies claims against trusted news sources
Performs AI-generated content likelihood analysis
Detects emotionally manipulative language patterns
Produces a summarized investigation report

The extension displays:

Credibility score
AI-generation likelihood score
Manipulation risk score
Key findings from the scan

How we built it

We built a Chrome extension popup interface using:

React
Tailwind CSS

The backend handles verification logic using:

Python + FastAPI
Gemini API for reasoning and summarization
News/search APIs restricted to trusted domains
Backboard.io for caching and models

Challenges we ran into

True AI-generated content detection is still an unsolved problem. We had to shift from “detection” to likelihood analysis using multiple signals, which was both more honest and more technically achievable. Verification pipelines can easily trigger multiple API calls per scan. Designing a caching strategy early using Blackboard prevented runaway API usage.

Connecting:

content scripts
popup UI
backend endpoints

Required careful coordination between team members to ensure consistent data structures.

Accomplishments that we're proud of

We’re especially proud of:

Building a complete browser-extension verification workflow.
Implementing the content-hash caching to reduce API calls.
Designing a cohesive noir-themed investigation interface
Creating a structured credibility-analysis pipeline
Producing clear, explainable scan reports

What we learned

We learnt that:

Caching and system design matter as much as machine learning
Browser extension architecture requires careful separation of responsibilities
Misinformation detection is more about verification workflows than AI models alone