Making AI Writing Sound Like You Actually Wrote It
AI writing gives itself away through a handful of overused words and a flat, generic tone. This page breaks down how to build a banned-word list and a voice reference so your drafts read like a real person. It is the longer version of what I covered in the reel.
If a human would not say it out loud, cut it from your AI drafts.
Build a banned-word list once and make the AI check every draft against it.
Feed the AI your own past writing so it copies your tone, not a generic one.
Detection is not about clever tricks. It is about sounding specific instead of safe.
Edit the last 10 percent by hand. That is where your voice actually lives.
As promised in the reel
Get the doc, prompts and workflow.
The exact thing, ready to use. Copy it, change the parts in [BRACKETS], go.
Document
Banned AI-word list
In the reel: “Comment "WORDS" and I'll send you my full banned-word list”
A ready reference of overused AI words and phrases to strip from any draft before you publish.
# Banned AI-Word List
If a human wouldn't say it out loud, cut it. Remove or rewrite any of these before publishing.
## Overused single words
- Delve
- Unlock
- Seamless / seamlessly
- Elevate
- Leverage
- Robust
- Harness
- Navigate (figurative)
- Tapestry
- Realm
- Myriad
- Foster
- Streamline
- Empower
- Cutting-edge
- Pivotal
- Crucial
- Vibrant
- Bustling
## Overused phrases
- Game-changer / game-changing
- In today's fast-paced world
- In the ever-evolving landscape of
- At the end of the day
- When it comes to
- It's important to note that
- Dive deep into
- Take your [X] to the next level
- Unlock the power of
- A testament to
- Pave the way
- The world of
## Stylistic tells to strip
- Starting sentences with "Moreover," "Furthermore," "Additionally"
- Over-balanced "not only... but also" constructions
- Three-item lists for everything
- Summary paragraphs that start with "In conclusion" or "Overall"
Add new words here whenever you catch your AI repeating itself.
How to use it: Keep this list handy and paste it into your AI tool, or scan your own drafts against it. Add your own words as you spot them.
Rebuilt from what the reel shows. Swap in your own details.
Prompt
Voice-matching and word-removal prompt
In the reel: “Just write voice below and I'll share the exact details on what I just described”
Makes AI write in your own tone by learning from your past posts, then scrubs the AI tells out.
I want you to write in my voice, not generic AI voice.
First, study how I actually write. Here are examples of my existing posts and copy:
[PASTE 3 TO 5 OF YOUR OWN POSTS OR WEBSITE SECTIONS HERE]
From these samples, extract my tone of voice: sentence length, rhythm, how formal or casual I am, words and phrases I actually use, and how I open and close.
Now take the following draft and rewrite it in that exact tone of voice:
[PASTE THE DRAFT YOU WANT REWRITTEN HERE]
Rules:
1. Match my voice from the samples above, not a neutral AI style.
2. Remove every word and phrase from this banned list: [PASTE YOUR BANNED-WORD LIST HERE].
3. If a human wouldn't say it out loud, cut it or replace it with plain words.
4. Keep my meaning intact. Do not add filler, summary lines, or transition words like "moreover" and "furthermore."
Before you give me the final version, list any banned words you found and what you replaced them with.
How to use it: Paste into your AI tool. Drop in several of your real posts or website copy where shown, then paste the draft you want rewritten.
Rebuilt from what the reel shows. Swap in your own details.
Prompt
Draft-checking prompt against your banned list
In the reel: “Have it check every draft against them before you publish”
Scans a finished draft for AI tells and flags them so you can fix before publishing.
Check the draft below against my banned-word list before I publish.
Banned words and phrases:
[PASTE YOUR BANNED-WORD LIST HERE]
My voice rules:
- [RULE 1, e.g. short sentences]
- [RULE 2, e.g. no corporate jargon]
- [RULE 3, e.g. write like I talk]
Draft:
[PASTE YOUR DRAFT HERE]
Do this:
1. Flag every banned word or phrase you find and where it appears.
2. Flag any sentence that sounds like AI even if it isn't on the list.
3. Suggest a plain-language replacement for each.
4. Give me a corrected version of the full draft.
How to use it: Paste into your AI tool with your banned-word list and the draft you're about to publish.
Rebuilt from what the reel shows. Swap in your own details.
Workflow
Three ways to make AI writing sound human
In the reel: “There's like three ways to do this”
The three approaches Fritz uses to strip AI tells and write in your own voice.
Three methods for removing AI tells from your writing.
1Method 1: Build a skill file. Pull a reference list of overused words (from a Wikipedia or GitHub page) and have your AI check your draft against it so those words never get used.
2Method 2: Copy an existing skill file. Grab one someone else already built on GitHub and plug it into your AI as the reference list.
3Method 3: Build your own voice. Collect your existing posts and website copy, feed them to the AI, tell it to match that tone, then have it remove the banned words from every draft.
4Run every draft through your chosen method before publishing.
How to use it: Pick one approach or combine them. The third is the most personal.
Rebuilt from what the reel shows. Swap in your own details.
Going deeper
The words that give it away
Certain words are fingerprints. Delve, unlock, seamless, game-changer, in today's fast-paced world. No one says these in a normal conversation, but AI reaches for them constantly because they are statistically safe.
The fix is boring and it works. Keep a running list of the words you keep catching, and tell the AI to never use them.
My read is that the list matters less than the habit of maintaining it. Every time a draft makes you cringe, add the offending word. After a month you have something personal.
Why banning words is not enough
Cutting the obvious words only removes the surface tells. The deeper problem is tone. AI writes in a flat, agreeable middle voice that never commits to anything.
Human writing has edges. Short sentences that land. A strong opinion stated without hedging. A specific example instead of a vague claim.
I keep seeing people run their text through a humanizer tool and expect magic. Grammarly, Scribbr, and ChatGPT all offer one, and they help with phrasing. But they cannot give you a point of view. That part is still your job.
Teach the AI your actual voice
Paste in three or four things you have already written. Old posts, a website section, an email you were proud of. Then tell the AI to study the rhythm and match it.
This works far better than asking for a vague "casual tone." The AI needs real samples, not adjectives.
A grounded example: I took five of my own captions, dropped them into Claude, and asked it to pull out patterns. It noticed I open with statements, use hyphens over commas, and keep paragraphs to two lines. Now every draft starts closer to how I already sound.
The three ways to set this up
First, build a skill file. This is a reference document the AI checks every draft against, pulling flagged words from public lists on GitHub or Wikipedia.
Second, borrow someone else's file. Plenty of people share theirs openly, and it is a fast starting point before you customize.
Third, build from your own writing. Reference your existing posts, extract the tone, then layer the banned-word list on top. This last one takes the most effort and produces the most convincing result.
The last pass is always human
No tool catches everything. The final edit still comes down to reading your draft out loud and fixing what sounds off.
If a sentence makes you pause because it sounds like a brochure, rewrite it in the words you would use talking to a friend.
I think the whole point gets missed when people treat this as a trick to beat a detector. The goal is not to pass a scan. It is to write something a person actually wants to read, which happens to be the same thing that sounds human.
Want to go deeper?
Let's look at it together.
One call, one on one. We take this and apply it to your account or your workflow.