TrustLens

Inspiration

Few days back i was signing up in one site and i accept the terms and condition without reading it as always i do. its now in my muscle memory that i always agree the terms and condition check box and continue , that time i started wondering how many important things we are missing because we don't read these polices . and i am not alone most of people are ignoring it just because it is lengthy ,complicated and most of us not have time to read them . That when i decided to build something for people where just in one click they can see the important information so they can decide if they want to continue before clicking "agree" . and finally i select to build an chrome extension as i think it is the fastest and most convenient way to make this .

What it does

TrustLens is a chrome extension that help people to understand what they are actually agree before signing up or making payment to any site. User just have to click once and extension will fetch all important policy pages like Terms & Conditions, Privacy Policy, Refund Policy, Cancellation Policy, and more . it reviews all of this using powerful ai model and highlights only the important clauses so user don't have to read thousands of lines of legal texts.

How we built it

TrustLens is built as a Chrome extension with a custom website crawler that automatically discovers policy pages across different websites. We used o4-mini/gpt-os-120B model through the OpenAI API/cloudflare api to analyse legal documents, extract the important clauses, categorize them, compare policy versions, and generate meaningful change summaries. everything will be stored in local including History, setting and watchlist. A huge part of the project was built using Codex. We used Plan Mode before writing code to design the architecture and implementation plan, and Goal Mode to discover and fix edge cases automatically. Throughout development we mainly used GPT-5.6 (Terra) with High reasoning, and switched to Extra High and Ultra reasoning for difficult architecture decisions, crawler issues, and debugging.

Challenges we ran into

The Biggest challenge was every website add the legal pages differently like some websites directly have in footer , some hided inside some menu , some also added in different sub domain and i also see some new startup sites don't even link to anywhere .Some are static and others are js rendered. So we have cover all of this and able to scan all this types of site. Another challenge was regarding the Policy watch feature. simply comparing the text produce too many false positive because small wording or grammar changing happen frequently. So we have give lots of time in prompt refinement in such way that if only important information is changed then only user should get notified . handling this things above things with keep extension experience simple was the challenge.

Accomplishments that we're proud of

We're most proud of Policy Watch, which turned into one of the core features of TrustLens. Instead of making users manually check policies every few months, it automatically detects important changes and explains exactly what changed.

We're also proud that every AI-generated claim is linked back to its original source, making the results transparent instead of asking users to blindly trust the AI.

Finally, we're happy that we built the entire project—from planning and architecture to the final UI polish—using Codex and GPT-5.6 while keeping the extension clean, simple, and easy to use.

What we learned

One of our biggest lessons was that spending more time planning actually saves time later. Using Plan Mode before writing code helped us avoid major architecture changes during development.

We also learned that AI performs much better when it's given enough context instead of isolated tasks. Goal Mode helped us uncover edge cases that we probably wouldn't have found ourselves.

Most importantly, we realized that building a reliable AI product isn't only about prompts or models. The real work is handling real-world edge cases, validating outputs, and making sure users can trust every result.

What's next for TrustLens

we are planning to support on more browsers and in more languages and make more accuracy in policy analyser and change detection. we also want to introduce the hosted version along with local version of the extension so users don't have start the server in local and works even when their laptop is off . so in short In the long term, we want TrustLens to become a tool that helps millions of people stay informed about what they are aggreging and keep updated is any change occurs,

Built With

• chrome
• extension
• manifest
• node.js
• react
• typescript
• vite