This project addresses the challenge of generic and verbose AI-generated academic writing, offering a specialized tool to refine drafts for scholarly papers and grant proposals. It targets researchers and academics seeking to maintain their authentic voice and scholarly precision while leveraging AI assistance. The primary benefit is the production of clearer, more concise, and author-aligned academic prose that accurately reflects claims and evidence.

How It Works

The tool employs a sophisticated six-layer processing pipeline: a general AI-tell catalog, academic-specific tells, preservation of scholarly conventions, claim-evidence matching, voice/venue calibration, and a dedicated funding-proposal mode. This audit-rewrite loop is designed to meticulously refine AI-assisted drafts. Its novelty lies in its calibration against accepted papers and funded proposals, focusing on sharpening precision and aligning with the author's unique style, rather than generic text "humanization."

Quick Start & Requirements

• Install: Clone the repository: git clone https://github.com/AIScientists-Dev/academic-humanizer ~/.claude/skills/academic-humanizer
• Prerequisites: Functions as a "skill" or system prompt for AI agents like Claude Code, Codex, and MorphMind. Forking the repository allows adaptation to personal writing styles and fields.
• Links: Examples are available at examples/before-after.md. The core logic is defined in SKILL.md.

Highlighted Details

• Sharpens clarity and voice by removing common AI phrasing (e.g., "In recent years...", "paves the way") and lengthy sentences.
• Preserves essential scholarly elements like evidence-based hedging, passive voice where appropriate, definitions, symbols, and all citations.
• Ensures claims are rigorously tied to evidence, replacing overly strong verbs (e.g., "prove") with data-supported alternatives (e.g., "show empirically").
• Features a distinct mode for NSF/NIH grant proposals, prioritizing structure and first-page impact crucial for reviewer scoring.

Maintenance & Community

Developed with support from NSF, CAREER, and NIH R01 grants. It builds upon the general AI-tell catalog from blader/humanizer and is complementary to broader projects like koaeraser/ARMS. No direct community links (e.g., Discord, Slack) are provided in the README.

Licensing & Compatibility

• License: MIT.
• Compatibility: The MIT license permits broad use, including commercial applications and integration into closed-source projects.

Limitations & Caveats

This tool serves as an editing aid and is not designed to evade AI-use detection systems. Users are responsible for adhering to venue-specific AI disclosure policies. Effective use requires personalization; forking the repository and adapting its rules to match an individual's voice and field is recommended. Direct execution often relies on integration with specific AI agent platforms.