Harvard Student Builds Tool to Make AI-Generated Writing Look Human
Ben Horwitz, a Harvard Business School student, has launched Sincerely, a Chrome extension that deliberately introduces errors and casual phrasing into polished text. The tool reverses the logic of grammar checkers like Grammarly, adding imperfections instead of removing them.
Horwitz created the tool to address a practical problem: readers now assume flawless writing comes from ChatGPT or other AI systems. When humans produce grammatically perfect emails, colleagues mistake them for machine-generated content.
Why Perfect Writing Now Signals AI
Large language models trained on billions of words produce text that is coherent, error-free, and contextually appropriate. This capability, once the mark of skilled writing, is now common enough that readers treat it with suspicion.
Professional writers report frustration on platforms like LinkedIn. They say years of developing craft now work against them-their polished prose gets flagged as AI-generated. AI detection tools compound the problem by flagging human writing that happens to match patterns the models learned from.
The result: some writers now intentionally include typos and looser phrasing to signal they're human.
How Sincerely Works
The tool offers three modes:
- Subtle: Removes unnecessary words and converts phrases to contractions
- Human: Adds conversational tone and minor typos in opening sentences
- CEO: Converts text to lowercase, shortens it drastically, and optionally adds "sent from my iPhone"
The extension is free for limited use, then requires a $4.99 subscription.
Horwitz's Personal Motivation
Horwitz has mild dyslexia and types slowly. Early in his career, tools like Grammarly helped him write faster and more correctly. Now his inbox overflows with perfectly polished AI-generated messages, and he wanted to restore authenticity to digital communication.
He built Sincerely using Claude, an AI model, to solve the paradox: using AI to make writing look less like AI.
What This Means for Writers
AI for Writers has created a new problem. Readers can no longer trust their instinct that good writing signals human effort. Detection tools fail because human writing naturally resembles the patterns AI models learned from.
For professional writers, the challenge is no longer perfecting prose-it's proving the prose is yours.
Your membership also unlocks:
