Ghost Writer: an AI writing tool with built-in quality checks
Ghost Writer drafts articles and posts from patterns studied in strong published writing, checks every draft against a 40-point QA list and revises what fails. It then formats the piece for the platform it is going to.
- Product
- Ghost Writer
- Built
- Mar 2026
- Code
- Public on GitHub
- Stack
- Node.js, Cloudflare Workers, LLM APIs
- Output
- LinkedIn, Reddit, Substack, Docs, Word and more
The idea
Most AI drafts read the same: stock phrases, paragraphs of equal length, lists of three and a conclusion that repeats the intro. Editing that out afterward takes longer than it should.
We wanted a tool that writes well from the first draft and checks its own work before a person edits it.
How we built it, step by step
- Studied strong writingWe drew structural patterns from 40+ articles in outlets such as HBR, ESPN, CNN, WSJ and Search Engine Land, and from 50+ papers on writing and AI text.
- Built a 5-stage pipelineEach piece moves through profile intake, pattern injection, 40-point QA, platform adaptation and polish.
- Added the 40-point QAThe checks cover linguistic patterns, readability scoring, bias and governance, among other groups. When a draft fails, a targeted revision loop fixes the weak parts instead of rewriting everything.
- Flagged generic phrasingA list of 200+ overused phrases and 8 structural pattern checks catch filler before it reaches an editor.
- Defined voice profilesVoice profiles describe structural habits, like the spread of sentence lengths and how paragraphs are built, rather than catchphrases. You can import your own.
- Formatted for each platformOutput adapters format drafts for LinkedIn, Reddit, Substack, Google Docs, Word and more.
- Hosted the web appThe app on ahmeego.com adds grammar checking and readability scoring, and the repo includes a CLI and a 10-chapter wiki.
What it does
- 5-stage writing pipeline
- 40-point QA with targeted revisions
- Overused phrase and structure checks
- Voice profiles based on structure
- Formats for LinkedIn, Reddit, Substack, Docs, Word and more
- Grammar and readability scoring
- CLI and hosted web app
How people find it
- Hosted app on ahmeego.com
- Public source on GitHub
- CLI: ghost write --type
- 10-chapter wiki
Questions people ask about this work
How do I make AI-written content less generic?
Give the model structure to follow, not just a topic, then check the draft for stock phrases and repetitive patterns. Ghost Writer builds drafts from patterns found in strong journalism and runs a 40-point QA before you edit.
What phrases make writing sound AI-generated?
Common signs are stock openers, summary conclusions, lists of three and paragraphs of the same length. Ghost Writer checks drafts against 200+ overused phrases and 8 structural patterns and revises what it finds.
Can AI write in my brand's voice?
It can get close if you describe the voice by structure: sentence length, paragraph shape and how you open and close. Ghost Writer stores those habits as a voice profile, and a person should still edit the final draft.
Services in this project
Related projects
ContextOS
ContextOS is a Python MCP server and CLI that gives AI agents memory, routes retrieval by how fast data changes and adds a reasoning step between retrieval and generation.
AI Agents Crash Course
This free, MIT-licensed course takes learners from no experience to deploying multi-agent systems in 5 days.
Buddy by Ahmeego
Buddy is a chat-based agent that reads a Google Ads account, stages changes, asks you to confirm them, makes them through the Google Ads API and checks the result.
Built and written up by John M. Williams, founder of It All Started With An Idea in Queen Creek, AZ. 15+ years and $350M+ in managed ad spend.