Schools and workplaces now run almost everything through AI detectors, and people are getting flagged for writing that is genuinely their own. Until that settles down, the fix is simple: your writing just has to actually read like a person wrote it. This is the exact setup I use to make Claude do that. It takes about 30 seconds to set up once, and then it is one command forever.

Every prompt below has a Copy button. Do these four steps in the same Claude chat, in order.

1. Tell Claude to humanize everything

Start a new chat and set the rule for the whole conversation.

The prompt
I want you to humanize all my text moving forward so it never sounds like AI. Write the way a real person actually writes: vary your sentence length, use plain words and contractions, drop the corporate tone, and never pad. From now on this applies to everything you write for me.

2. Feed it the actual list of AI tells

There is a real Wikipedia page called Signs of AI writing (en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing) that catalogs every pattern that gives AI away. Hand it to Claude and make it swear off all of them.

The prompt
Look up the Wikipedia page titled "Signs of AI writing" (Wikipedia:Signs of AI writing). List out every single pattern on it that gives away AI generated text. Then commit to never doing a single one of them again in anything you write for me.

For reference, the biggest tells it will find are em dashes everywhere, the rule of three (everything comes in lists of three), filler phrases like "in today's fast-paced world" and "it's important to note", inflated words like delve, tapestry, testament, and unlock, empty hedging, and the "it's not just X, it's Y" construction. Once Claude sees the full list, it stops doing them.

3. Save it as a /human skill

Here is the trick that makes this permanent. Stay in the same chat, open the slash menu, choose skill-creator, and have it package everything you just did into a reusable skill.

The prompt
/skill-creator Build me a personal humanizer skill from everything we just covered, so any time I type /human you rewrite in a natural human voice that avoids every AI tell we listed. Call the skill "human".

4. Install it and use it

When Claude finishes building the skill, press Save skill to install it. That is the whole setup, and you never have to explain any of this again. From now on, just put /human at the start of any prompt and Claude writes in your natural voice.

How you use it
/human write me a cover letter for this job: [PASTE THE JOB]

The all-in-one prompt (if you skip the skill)

Do not want to set up a skill? Paste this single prompt any time you need to clean up a piece of writing. It does the same job in one shot.

The humanizer prompt
Rewrite the text below so it reads like a real person wrote it. Kill every AI tell: no em dashes, no "in today's world", no "it's important to note", no forced lists of three, no inflated words like delve, tapestry, testament, unlock, or supercharge, no empty hedging, and no "it's not just X, it's Y". Vary sentence length, use contractions, cut filler, and keep my actual meaning and facts exactly. Return only the rewritten text.

Here is the text: [PASTE]
No detector is perfect, and this is not about faking anything. It is about getting AI to stop writing in the stiff, obvious way that gets real work flagged, so what you hand in sounds like you.

Two things that make it work better

  • Give it a sample of your writing. Paste a paragraph you actually wrote and tell it to match your voice. That is the single biggest upgrade.
  • Read it out loud before you send it. If a line sounds like a robot, tell Claude which one and it will fix it. Two passes and it is clean.

Set this up once and you will never think about it again. That is the whole point.
Anir

If you want this kind of AI actually built into your business, apply to work with me. And to catch the next drop, join the newsletter.