AI Lie Detector from Text Typing Speed

Inspiration

I’ve always been curious about how people unknowingly show signs when they lie — not just through body language but even while typing. I read about how typing speed, pauses, and corrections can indicate stress or hesitation. That gave me the idea to build a simple AI-based tool that detects lies based on how someone types their answers, not what they type.

What it does

This AI Lie Detector tool analyzes a person's typing speed, pause durations, and correction patterns while they type a response.

First, it records the person’s normal, honest typing behavior through a simple sentence. Then, it asks them to answer another question, where they might lie. By comparing both typing patterns, the tool predicts whether the person is likely being honest or deceptive — based on changes in typing speed, hesitation time, and corrections.

It doesn’t read the content of what is typed — it only observes how it’s typed.

How we built it

I created a web page where users type a simple, honest sentence to record their natural typing speed and style. Then, I asked them to answer another question, where they might try to lie. I captured their typing speed, pause time, and corrections in both cases. I built a simple logic to compare their normal behavior and response behavior to predict whether they were being truthful or not. I hosted the project live on Netlify for easy access.

Challenges we ran into

One big challenge was making sure the baseline typing pattern was accurate, because some people naturally type slower or make more corrections. Another issue was handling nervous typing, which can happen even when people tell the truth. Also, capturing real-time typing data smoothly on different devices and browsers needed extra testing.

Accomplishments that we're proud of

Successfully built a working AI tool that detects possible lies based on typing behavior, not the actual text.

Learned how to capture and analyze real-time typing metrics like speed, pauses, and corrections.

Created a clean, interactive demo that works smoothly in the browser.

Deployed the project live on Netlify, making it accessible to anyone instantly.

Turned a creative and unique idea into a functional prototype within limited time during the hackathon.

What we learned

While working on this, I learned how to capture typing patterns like speed, pauses, and backspaces in real-time. I also got to understand how keystroke behavior changes in different situations. I explored how to compare a person’s normal typing behavior with their response behavior to detect possible lies. This project also taught me about basic data classification and deploying a live app online.

What's next for AI Lie Detector from Text Typing Speed

Add a machine learning model that improves accuracy by training on larger, real-world typing behavior datasets.

Include emotional state detection by analyzing not just speed but typing pressure (if device supports it) and error frequency.

Create a multilingual version that works with regional languages and different typing patterns.

Build a mobile-friendly version so it can be used on smartphones and tablets.

Enhance the dashboard to show detailed typing analytics and visualizations like speed graphs and pause heatmaps.

Integrate it into chat apps or web forms for real-time behavioral analysis in sensitive or high

Built With

• chrome
• github-(for-code-hosting)-apis:-none-(for-now-?-but-can-integrate-ml-apis-in-future)-tools:-vs-code-(for-development)
• google
• javascript-event-listeners-(for-typing-pattern-capture)-platforms:-netlify-(for-hosting-the-live-demo)
• javascript-frameworks/libraries:-streamlit-(for-frontend-and-app-interface)
• languages:-python