• 

  Starting Page

• 

  One you choose a language for example I chose Arabic it will write everything and format everything that way.

• 

  This is the output

Inspiration

We kept coming back to one moment: someone moves to a new country, doesn't speak the local language fluently yet, and gets a panicked message in a community group chat, a rumor about a new immigration rule, a clinic closing, or a safety threat near their kid's school. They have no fast way to check if it's true, and most fact-checking tools only exist in English. We wanted to build something for that exact moment, for people who are already navigating enough uncertainty without adding misinformation on top of it.

What it does

Trust Line lets a user pick their language from a list that includes Arabic, Swahili, Urdu, Mandarin, Spanish, French, and more. From there, they can paste a message or upload a screenshot of one since most rumors show up as forwarded texts or images, not clean articles. Trust Line reads the claim, checks it against credible sources, and responds in the user's own language with a clear verdict: likely true, misleading, or unverified, along with a plain explanation and the sources behind that answer.

How we built it

We built the frontend using AI-assisted prompting to quickly scaffold the language selection, screenshot upload, and result screens, then connected it to the Claude API to handle the actual claim-checking logic, reading the text or extracted screenshot content, comparing it against available sources, and generating an explanation in the selected language. We kept the design simple on purpose: pick a language, paste or upload, get an answer you can actually understand.

Challenges we ran into

One of the biggest challenges was ensuring that the AI could provide accurate and transparent results rather than simple yes-or-no answers. We also faced challenges in processing different formats of information, especially screenshots containing text. Another challenge was balancing speed, accuracy, and source reliability while maintaining a smooth user experience.

Accomplishments that we're proud of

We're proud that Trust Line isn't built for "everyone" in the abstract, it's built for a specific person dealing with a specific kind of stress: language barriers plus misinformation. We're also proud we got multilingual support and screenshot handling working in the same flow, since most teams building something like this in a week would only manage one or the other.

What we learned

We learned that translation and fact-checking are actually two separate hard problems, not one, getting an accurate verdict is only half the job if the explanation doesn't land clearly in someone's first language. We also learned a lot about where AI should stop and a human should take over, especially for anything involving legal status or safety.

What's next for Trust Line

We'd want to expand the language list further, add a "verify with a local organization near you" directory, and explore a lightweight version that works over SMS for users with limited data access, since many of the people we built this for may not have reliable smartphone internet.

Built With

• and-image-text-extraction.-we-used-anthropic's-native-web-search-tool-for-real-time-source-verification
• anthropic
• claudeapi
• css
• fact-checking
• html
• javascript
• node.js
• react
• vercel