🚨 AI is powerful — but not always trustworthy.
TrustLens helps students evaluate whether an AI-generated answer is safe to submit, cite, or rely on.
TrustLens shifts AI usage from blind trust to informed decision-making.
🌐 Live App: https://trustlens-tano.streamlit.app/
🎥 Demo Video: (https://www.youtube.com/watch?v=UFmjlNo5Mrw)
📦 Repository: https://github.com/eliza-ochoa/TrustLens
TrustLens is a deterministic Streamlit app that helps students judge whether an AI-generated answer is trustworthy enough to submit, cite, or revise.
Instead of generating more content, TrustLens evaluates existing AI outputs using transparent, rule-based logic and provides actionable insights.
AI-generated answers often sound confident and complete, even when they are unsupported or misleading. Students frequently rely on these responses without verifying them.
TrustLens introduces a lightweight, explainable review step that helps users detect risk signals, evaluate evidence alignment, and revise unsafe outputs before using them in academic or real-world contexts.
TrustLens focuses on fast, transparent evaluation rather than content generation.
- 📊 Prominent Trust Score with High, Medium, or Low Trust badge
- 🔎 Clear review sections:
- Top Concerns
- What an evaluator would expect to see
- Safer Rewrite
- Audit Trace
⚠️ Deterministic rule-based checks:- unsupported certainty
- vague attribution
- missing source support
- overgeneralization
- mismatch between answer and source text
- unsupported numbers, dates, and named claims
- 🧠 Plain-English Explainability section
⚠️ Validation Boundaries section- 🛡️ Responsible AI banner + structured auditability
This score reflects rule-based risk signals and evidence alignment, not factual verification.
🎥 Demo Video: (https://www.youtube.com/watch?v=UFmjlNo5Mrw)
🌐 Live App: https://trustlens-tano.streamlit.app/
The demo shows how TrustLens evaluates a weak AI answer, highlights risk signals, and generates a safer rewrite in seconds.
| Home Screen | Low Trust Example | High Trust Example |
|---|---|---|
 |
 |
 |
This project was built and refined using Codex for:
- scaffolding
- debugging
- scoring logic refinement
- UI iteration
- final polish and documentation
TrustLens is designed to make responsible AI use visible in the product itself.
- Transparency – scoring is deterministic and rule-based
- Explainability – users can see how results were generated
- Safety – clear disclaimers and limitations are provided
- Validation Boundaries – no external or live web verification
- Auditability – traceable signals, rules, and score drivers
- Responsible Use – encourages verification against reliable sources
TrustLens demonstrates how responsible AI principles can be embedded directly into user-facing tools.
- Paste an AI-generated answer
- (Optional) Add source text
- Select a context
- Click Analyze Answer Trust
TrustLens will:
- assign a trust score
- highlight risk signals
- show expected evidence
- generate a safer rewrite
pip install -r requirements.txt
streamlit run app.py- Sentence-level evidence highlighting
- Optional file upload for source text
- Exportable review summary for classroom use
- More sample scenarios for live demos
- Students using AI for assignments or research
- Anyone who wants to validate AI-generated content before relying on it
- Users interested in responsible and transparent AI usage
TrustLens demonstrates how responsible AI principles can be embedded directly into user-facing tools.
TANO Research
Trust • Accuracy • Neutrality • Organization • Transparency
“Exploring Possibilities Everywhere”