NewsLens AI

• 

  NewsLens App Icon

Inspiration

In moments of local crisis—whether it’s a sudden rumor, a safety scare, or a neighborhood service disruption—misinformation spreads faster than truth. We noticed that when people are highly stressed and face an overwhelming amount of internet hype, they don't have the time or emotional bandwidth to dig through dozens of search results or verify what's real. We realized that people miss out on true, trusted guidance not because it isn't out there, but because scattered information and complex systems aren't designed for people under stress. We wanted to build a simple, supportive tool that gives users absolute clarity right when they need it most. They don't just need a list of web links; they need quick, calm, and highly organized clarity. This inspired us to create NewsLens an AI-powered assistant designed to turn chaotic local rumors into verified, structured reality.

What it does

Our project is an AI-powered Tracker built to help anyone evaluate local claims and understand what is true, misleading, or unverified. It provides a simple, accessible user interface (UI) built for everyone where they can input a local claim, piece of news, or rumor they heard. Instead of forcing the user to manually scroll through search results or click multiple links, the application uses Natural Language Processing (NLP) and AI to instantly cross-reference the claim across various news vendors and certified online sources. It checks the statement's accuracy against credible, accredited websites and outputs a clear verdict alongside direct links to trustworthy sources.

How we built it

We followed a clear User Input Output design flow to keep the platform highly accessible for everyday users. • The User Interface: Built a minimalist text-input system where a user simply types a claim or news headline they heard. • The Input: The user types a simple text query or rumor they heard into a clean, minimal user interface. • The AI Processing: Instead of executing a simple web search, our application uses Natural Language Processing (NLP) for text classification and summarization. The AI breaks down the user's claim, identifies the key subjects and queries public data sources and news vendors to find relevant reporting. • The Output: The system cross-references the findings, ranks source credibility, and displays a straightforward verdict, alongside links to accredited sites for complete transparency.

Challenges we ran into

Handling Nuance and "Satire" in News The Challenge: While building our tracker, we found that AI can easily struggle with context. For example, if a sports news site posts a satirical joke headline about an athlete, a standard AI might accidentally classify it as a flat-out lie or true fact, rather than recognizing it as humor. The Solution: We refined our AI prompt formatting to prioritize checking against accredited media databases rather than just scanning social media buzzword frequency, teaching the AI to evaluate the authority of the publishing source rather than just the words in the headline.

1. Balancing Automation with User Safety (The "Auto-Verify" Trap) The Challenge: We initially wanted the AI to definitively state "100% True" or "100% False" for every rumor. However, we realized that giving an AI absolute authority over the truth can lead to dangerous misinformation if the AI hallucinates. The Solution: We pivoted to a safer, more thoughtful approach, instead of auto-verifying with total certainty, our UI focuses on Source Transparency. The app flags the claim and immediately shows the user links to accredited sites so that a Human-in-the-Loop remains the final decision-maker alongside certainty range.
   
2. Avoiding "Feature Creep" Under Time Constraints The Challenge: During the brief build window, it was tempting to try and make our tracker check every type of local rumor—from school safety threats to public transit closures. We quickly realized that trying to cover everything made our database messy and diluted our app's effectiveness. The Solution: We decided to apply to highly relatable use case of media and local internet hype to cleanly demonstrate our user input output workflow.

Accomplishments that we're proud of

Building a Functional application in a Week: We are incredibly proud of taking a complex idea—combining AI processing with real-time news evaluation—and building a working prototype within the short hackathon window.

Creating an Intuitive, Stress-Free UI: We successfully designed a clean user interface that handles a messy problem (misinformation) and turns it into a clear, calming experience for the user.

Mastering the Balance of AI and Human Oversight: We didn't just build a blind automation tool; we successfully integrated the hackathon’s Human-in-the-Loop and Responsible AI requirements by ensuring our app focuses on source transparency rather than making absolute judgments for the user.

What we learned

We learned that trying to solve "misinformation" is not impossible for a small team in a short time. Focusing heavily on a relatable use case allowed us to prove our system design works perfectly. The Reality of AI Prompting: We learned how easily AI can be tricked by nuances like satire or internet hype, and we discovered how to refine our system logic to judge a source's authority rather than just reading headlines. Responsible AI is a Feature, Not an Afterthought: This project taught us that designing guardrails—like displaying accredited source links so users can verify data themselves—is just as important as writing the core code.

What's next for NewsLens AI

SMS and WhatsApp Integration: Stressed parents and busy caregivers don't always have the time to open a dedicated web app or browse a new interface. We plan to build a lightweight SMS and WhatsApp chatbot wrapper for NewsLens. Developing a chatbot that allows users to forward suspicious messages for immediate AI-assisted analysis directly in their chat. Multi-Language Support: Misinformation often targets communities where English is a second language, as official government updates can take longer to be translated. We want to integrate multi-language translation pipelines so that a user can input a rumor in their native language and get verification from local sources and receive translated summaries. Real-Time Web Scraping Integration: We plan to transition from our simulated news vendor dataset to live API integrations with real-time accredited fact-checking directories (like 211.org or global media registers). A "Flag for Review" Community Feature: To further enhance our Human-in-the-Loop model, we want to allow users to flag unverified rumors so that a community review panel can look into them, turning the app into a collaborative local safety net.

Built With

• base44
• chatgpt
• cursor
• gemini
• grok
• lovable