AI Writing Detector (ai-check)
Runs a forensic, evidence-cited scan of any text across nine AI-tell categories and returns a 0–27 score with a verdict and fixes.
What it does
- Scores text 0–3 on nine signal categories: word predictability, sentence-length uniformity (burstiness), hedge density, structural tells, specificity deficit, transition fingerprints, punctuation fingerprints (em dashes, semicolons, mid-sentence colons), voice/register, and rhetorical scaffolding.
- Converts the 0–27 total into a verdict (
Human → AI) plus a confidence level and a fixed report layout. - Separately estimates the AI-edited fraction (pure human → pure AI), which handles the common real case of a human editing an AI draft.
- Produces an evidence log where every fired signal must cite an exact quote and severity, and adds concrete rewrite suggestions when the score exceeds 6.
- Ships a very specific pattern library: banned vocabulary, "In conclusion" closers, double em-dash wrapping, announcement sentences ("What surprised me was…"), tricolons, chiasmus closers, Slack register collapse, templated email sign-offs.
Who it's for
- Writers, marketers and newsletter authors de-AI-ifying drafts before publishing
- Content editors humanizing LLM output at scale
- Teachers and reviewers who want a reasoned breakdown instead of a black-box percentage
- Anyone whose Slack updates or work emails read like a status report
Example uses
- Paste a blog draft: "run ai-check on this" → score, verdict, and a quote-by-quote evidence log.
- "What gives this LinkedIn post away as AI?" → it flags things like "Turns out…", "it's not X, it's Y", and symmetric tricolons.
- Check a team Slack update → catches the accomplishment → caveat → next-steps arc and overly well-formed sentences hiding under
lmkand~60%.
Caution: verdicts are heuristic, not proof of authorship. Confidence drops sharply for short texts, non-English writing, and iteratively paraphrased content — and some cited papers should be verified before quoting them.
· · · Install guide · · ·
Try it now, no install
Paste this into Claude to use the skill without installing anything.
Read the instructions in this file and follow them to help me: https://raw.githubusercontent.com/harshaneel/humanize/HEAD/plugins/humanize/skills/ai-check/SKILL.md What I want: (describe your task here)
If Claude can't open the link, open it yourself and paste the contents instead.
↓ If it works for you, download the ZIP below and install it. Then it runs on its own — no pasting each time.
Install in the Claude app (no terminal)
- Download the ZIP with the button below.
- In Claude, open Settings → Capabilities and turn on 'Code execution and file creation'. (one time)
- Go to Customize → Skills → + → 'Upload a skill' and upload the ZIP.
Install in Claude Code
Let Claude do it — paste this into Claude Code
Install the skill I found on Claude Skill Mart. Copy the plugins/humanize/skills/ai-check folder from the GitHub repo harshaneel/humanize into my ~/.claude/skills/harshaneel-ai-check/. When it's done, tell me in one line what this skill can do.
Install with a command instead
git clone https://github.com/harshaneel/humanize.git && mkdir -p ~/.claude/skills && cp -r humanize/plugins/humanize/skills/ai-check ~/.claude/skills/ai-check⚠ This is a third-party skill. Check the source repository before installing.
- Open a terminal (Terminal on macOS/Linux, PowerShell on Windows).
- Create the skills directory:
mkdir -p ~/.claude/skills - Clone the repo:
git clone https://github.com/harshaneel/humanize.git - Copy just this skill:
cp -r humanize/plugins/humanize/skills/ai-check ~/.claude/skills/ai-check - Verify that
~/.claude/skills/ai-check/SKILL.mdexists. - Restart Claude Code, then say "run ai-check on this" and paste the text you want analyzed.
- (Optional) For project-only use, copy the folder into your project's
.claude/skills/directory instead.