This preset provides textlint rules to detect and correct AI-generated writing patterns, promoting more natural Japanese expression. It targets technical writers, content creators, and anyone aiming to refine AI-assisted text. The primary benefit is improving the readability and naturalness of AI-generated content by focusing on structural rather than expressive constraints.

How It Works

The preset comprises several rules that identify common AI writing artifacts: unnatural list formatting, hype-driven expressions, mechanical emphasis, and English-centric colon usage. It employs morphological analysis for context-aware checks, particularly for colon usage, distinguishing between natural Japanese phrasing and AI-generated patterns. The design prioritizes structural guidance over strict expression rules, aiming for subtle improvements in naturalness.

Quick Start & Requirements

• Install with npm: npm install @textlint-ja/textlint-rule-preset-ai-writing
• Requires textlint 15.1.0 or later.
• Can be run as an MCP server with npx textlint --mcp for integration with AI tools like VSCode Copilot.
• Official documentation: textlint MCP documentation

Highlighted Details

• Detects and suggests alternatives for AI-specific list formatting (e.g., **重要**:, ✅).
• Flags overused hype expressions (e.g., "revolutionary," "game-changer") and provides more grounded alternatives.
• Identifies unnatural emphasis patterns (e.g., **非常に**重要な項目).
• Corrects English-style colon usage followed by block elements, promoting natural Japanese sentence structure.
• Includes ai-tech-writing-guideline for technical writing best practices (conciseness, active voice, specificity, consistency, structure).

Maintenance & Community

• Maintained by textlint-ja.
• Versioning follows semantic versioning (patch for fixes, minor for additions, major for breaking changes).
• Contributions via pull requests and issues are welcome.

Licensing & Compatibility

• Licensed under the MIT License.
• Compatible with commercial and closed-source projects.

Limitations & Caveats

The preset's rules, especially those related to AI-generated patterns, may occasionally flag natural language or require fine-tuning via the allows option. Rule modifications, even in patch versions, can alter error counts, necessitating the use of lock files for stable builds. The ai-tech-writing-guideline rule's effectiveness may depend on the specific technical domain.