Introduction

AI Guard demo — detecting AI-generated code patterns

What is AI Guard?

AI Guard is an AST-based code analyzer that detects patterns commonly found in AI-generated code. It helps maintain code quality by identifying verbose, over-commented, or poorly structured code before it reaches your repository.

Unlike traditional linters that focus on syntax and style, AI Guard specifically targets characteristics of LLM-generated code: excessive comments, generic variable names, obvious explanations, and AI-typical phrases.

The Problem

Large Language Models tend to produce code with predictable patterns that reduce maintainability:

• Excessive inline comments explaining obvious operations
• Generic variable names like 'data', 'result', 'item', 'temp'
• Verbose function documentation that restates the code
• AI-typical phrases like 'this function will...', 'we need to...'
• Unnecessarily long functions with redundant logic

AI Guard catches these patterns early, allowing you to refactor before committing.

AI Guard CLI usage output

Use Cases

• Pre-commit hooks to validate staged files
• CI/CD pipelines for pull request checks
• Editor integrations for real-time feedback
• Code review automation
• Educational tools for teaching clean code practices

Key Features

• AST-based analysis using ts-morph for accurate detection
• 5 built-in rules covering common AI patterns
• CLI tool with git staging support
• Programmatic API for custom integrations
• Configurable thresholds and rule settings
• Fast analysis with TypeScript and JavaScript support