The Kill List: 20 Words to Ban from Your ChatGPT Prompts Immediately

You know it when you read it. That slightly-too-formal, slightly-too-polished, weirdly-corporate writing that shows up in blog posts, LinkedIn updates, email newsletters, and social captions all over the internet right now. Nobody sat down and wrote it. An AI did. And the giveaway isn't that it's bad — it's that it uses the exact same AI words to avoid. Every. Single. Time.

I keep a kill list. It lives in every Claude Project I set up, every ChatGPT custom instruction I write, and every piece of content that leaves ConnectCraft. Here are the 20 words at the top of it — and why each one is dead.

Why AI Keeps Using the Same Words

Before the list, a quick explanation — because knowing why AI does this makes it easier to fix.

AI models are trained on enormous amounts of text from the internet. And a huge portion of that text came from formal sources: academic papers, press releases, corporate about pages, business blogs. These sources skew heavily toward formal, elevated, "professional-sounding" language.

So when you ask ChatGPT to write something, it pulls from that training data and gravitates toward the most statistically common words for the context. The safe center. The average. Which, in business writing, means a very specific cluster of words that now signal AI to anyone paying attention.

AI doesn't pick these words because they're good. It picks them because they're the most likely next word based on everything it's ever read. That's it.

The fix isn't complicated. You give AI a list of words it's not allowed to use — and it's forced to find other ways to say the same thing. Usually better ways. Here's your list.

The Kill List: 20 AI Words to Avoid in Every Prompt

01. DELVE

Why it's dead: This is the single biggest ChatGPT tell on the internet right now. One chart tracking word usage online shows "delve" skyrocketing in 2022 — the exact year ChatGPT launched. Nobody actually says "let's delve into that" out loud. Nobody.
Say instead: "Look at." "Dig into." "Break down."

02. LEVERAGE

Why it's dead: It means "use." That's it. AI uses "leverage" when "use" is right there, free, doing the same job with three fewer syllables.
Say instead: "Use." Just say use. Seriously.

03. TAPESTRY

Why it's dead: A tapestry is a woven fabric. It is not a metaphor for your business strategy, your content ecosystem, or human experience.
Say instead: "Combination," "mix," "range," or just describe the actual things.

04. SEAMLESS

Why it's dead: Nothing is seamless. Every tool has friction. When AI calls something "seamless," it's signaling that it has no actual experience using the thing it's describing.
Say instead: "Smooth," "simple," "easy to set up," or be honest about the friction.

05. ROBUST

Why it's dead: This word has been stripped of all meaning. A "robust content strategy" tells me nothing. Robust compared to what? It's a filler adjective that sounds like it means something but doesn't.
Say instead: "Strong," "solid," or describe what actually makes it good.

06. COMPREHENSIVE

Why it's dead: AI calls everything comprehensive because it doesn't know how to say "this covers a lot of ground" in plain English. Bonus flag: "comprehensive guide" in a headline is almost always written by AI.
Say instead: "Complete," "full," "covers everything from X to Y."

07. HARNESS

Why it's dead: You harness a horse. You don't harness AI capabilities, harness your potential, or harness the power of content marketing.
Say instead: "Use," "put to work," "apply."

08. EMPOWER

Why it's dead: Corporate therapy language. When something "empowers" you, it means it helps you do something. Just say it helps you.
Say instead: "Help," "give you the tools," "make it easier for you to..."

09. PIVOTAL

Why it's dead: Everything in AI content is pivotal. When everything is pivotal, nothing is. It's an inflated adjective that's doing the job "important" could do for free.
Say instead: "Key," "critical," "important," or just make the case for why it matters.

10. FURTHERMORE

Why it's dead: Nobody uses "furthermore" in conversation. Ever. It's a formal academic transition that signals immediately: this was not written by a human who talks to other humans. Same applies to "moreover," "additionally," and "consequently."
Say instead: "And," "also," "on top of that," "here's the other thing."

11. LANDSCAPE

Why it's dead: The "AI landscape." The "content landscape." AI uses this word constantly to sound like it understands the big picture without actually describing anything specific.
Say instead: "Space," "market," "industry," "what's happening in [your field]."

12. UNLOCK

Why it's dead: You unlock a door. You don't unlock your potential, your content strategy, or your business growth.
Say instead: "Get access to," "open up," "start using."

13. TRANSFORMATIVE

Why it's dead: Everything AI writes about is transformative. If you can't describe what actually changed and for who, this word is covering up a lack of specifics.
Say instead: Describe the actual result: "cut my prep time from 3 hours to 40 minutes."

14. IN TODAY'S EVER-EVOLVING LANDSCAPE

Why it's dead: The grand slam of AI tells. If a piece of content starts with these words, you can stop reading. No human opens a conversation this way.
Say instead: Just start with the interesting thing. Skip the scene-setting entirely.

15. IT'S IMPORTANT TO NOTE

Why it's dead: Why is it important? If it's important, just say it. This phrase is throat-clearing — AI's way of signaling that a significant point is coming instead of just making the point.
Say instead: Delete it. Say the thing directly.

16. SYNERGY

Why it's dead: A word that meant something in the 1990s boardroom and has meant nothing since. AI drags it out whenever two things work well together instead of just saying those two things work well together.
Say instead: Describe what actually connects: "when you combine X and Y, you get Z."

17. REALM

Why it's dead: "In the realm of content marketing." You are not writing a fantasy novel. There are no realms.
Say instead: "In content marketing." "With AI." No realm required.

18. METICULOUS / METICULOUSLY

Why it's dead: AI uses this word to signal that something was done carefully — without actually showing any of those things. It's a confidence claim with no evidence.
Say instead: Show the care through specifics: "we check every link, every stat, every claim."

19. FOSTER

Why it's dead: "Foster meaningful connections." "Foster growth." What does fostering actually look like? AI doesn't know, so it uses this word instead of saying.
Say instead: "Build," "create," "encourage," "develop."

20. GAME-CHANGER

Why it's dead: Everything is a game-changer. Every tool, every strategy, every framework. The word has been used so many times it now signals the opposite of its intent.
Say instead: Just describe what actually changed. Specifics are the real game-changer.

Bonus Round: The Phrases That Are Just as Bad

• "In today's rapidly evolving world..." — No piece of content that a human would actually want to read has ever started this way. Delete on sight.
• "It's worth noting that..." — If it's worth noting, note it. The meta-commentary adds nothing.
• "Let's dive in." — The official farewell to the intro paragraph. Every AI-written piece ends its intro with this.
• "As an AI language model..." — If you're seeing this in your content, something has gone very wrong.
• "In conclusion..." — Your conclusion doesn't need to announce itself. Just wrap up.

How to Actually Use This List

Having a kill list is one thing. Making it work every session is another. Here's the system.

Option 1: Bake It Into Your Prompt

Add a kill list section to the prompt you use every time you ask AI to write content. After your main prompt, add: "Never use any of the following words: delve, leverage, tapestry, seamless, robust, comprehensive, harness, empower, pivotal, furthermore, landscape, unlock, transformative, synergy, realm, meticulous, foster, game-changer, or the phrase 'it's important to note.'"

Option 2: Save It in Claude Projects or ChatGPT Memory

If you're using Claude, drop the kill list into your Project Instructions so it loads automatically every session. If you're using ChatGPT, add it to your Custom Instructions. You set it once. It runs every time.

Option 3: Use It as an Editing Checklist

Before you publish anything AI-generated, do a quick Control+F for the top offenders: delve, leverage, tapestry, seamless, robust, comprehensive. If any of them are there, replace them. Takes 90 seconds and immediately improves the output.

The Brand Voice Workbook at ConnectCraft walks through exactly this process — building your personal kill list, your voice reference doc, and your prompt system so you're not reinventing this every time you open a new chat. Grab it free at brandvoice.connectcraftai.com.

The Bigger Point

A kill list is a start. But it's not the whole fix.

The reason AI content sounds generic is that it doesn't know how you write. It doesn't know your rhythm, your humor, your specific opinions, or the words you actually use in conversation. So it defaults to the average of everything it's ever read.

The kill list removes the worst offenders. But loading your real voice — examples of your actual writing, your banned words AND your preferred phrases, the specific things you always say and never say — is what takes AI content from "not embarrassing" to "sounds genuinely like me."

AI isn't the problem. It's how you're using it. And a kill list, combined with a real voice training system, is how you fix it.

Frequently Asked Questions

What are the most common AI words to avoid?

The top offenders that appear most consistently across AI-generated content are: delve, leverage, tapestry, seamless, robust, comprehensive, harness, empower, pivotal, and furthermore. These show up so frequently in ChatGPT output that readers now associate them immediately with AI writing.

Why does ChatGPT use the same words over and over?

AI models predict the statistically most likely next word based on their training data. Most training data comes from formal sources — academic papers, press releases, corporate sites — that overuse these terms. So ChatGPT defaults to them because they're the safest, most common choice, not because they're the best choice.

How do I tell ChatGPT to avoid certain words?

Add a "do not use" list directly in your prompt. Example: "Do not use any of the following words: delve, leverage, robust, seamless, furthermore..." You can also add this to ChatGPT's Custom Instructions or a Claude Project so it applies automatically to every session.

Will banning these words make my AI content sound human?

Banning the worst offenders is step one. But for content that actually sounds like you, AI also needs your real voice examples — your best past content, your preferred phrases, the things you'd never say. The kill list removes the obvious tells. Your voice examples replace them with something real.

What's the fastest way to check if my content sounds like AI?

Do a Control+F for "delve," "leverage," "tapestry," and "furthermore." If any of them are there, you have AI fingerprints. Then read the piece out loud — if it sounds like you're presenting at a corporate conference instead of talking to a person, it needs a rewrite.