12 AI Tells and the Skills That Fix Them

I opened LinkedIn this morning and scrolled for a solid five minutes before I found a single post that sounded like an actual human wrote it. Everything else read like it fell out of a chatbot with zero editing. So when I came across this breakdown from a sharp AI professional, I stopped scrolling and actually read the whole thing twice.

The author’s point is blunt and I love it: nobody edits AI-written content anymore, and it shows. They collected 12 dead giveaways that scream “a robot wrote this” plus 11 Claude skills they personally use to scrub those tells out. I was nodding the entire way through, because half of these I’ve caught myself skimming past without even realizing why.

Let me break down what this LinkedIn creator shared, because it’s genuinely useful whether you write posts, emails, or reports.

The 12 AI tells to delete today

These are the words and patterns the expert says you should hunt down and kill in your own writing. Read them out loud. If you’d never actually say it to a friend, that’s your answer.

1. “Delve.” You’ve never said this word out loud in your life. Not once. Yet AI reaches for it constantly.
2. “Crucial / pivotal.” Almost nothing is that pivotal. The word inflates the stakes and fools no one.
3. “Tapestry.” As the author puts it, no human speaks like this. Ever.
4. “Here’s the thing.” The original poster nails it: there is no thing. It’s filler pretending to be insight.
5. “Hope this helps.” It doesn’t help, and worse, it outs you as a copy-paste job.
6. “After careful consideration.” Usually written without any actual consideration behind it.
7. “To provide a quick update.” Just give the update. Skip the runway.
8. “Most people…” The lazy oversimplification that starts a thousand generic takes.
9. “Robust / seamless / realm.” The corporate AI vocabulary starter pack.
10. Adverb abuse. Things that “quietly run” or “seamlessly integrate.” Nothing actually runs.
11. “It’s not X, it’s Y.” The creator calls this the most famous AI sentence alive, and once you see it you can’t unsee it.
12. “This isn’t a budget. It’s a statement of intent.” That dramatic reframe formula. Just no.

Why does this matter so much? Because trust is the whole game. The moment a reader spots one of these tells, they mentally file your writing under “machine wrote this,” and everything after that gets discounted. Sounding human is not a style choice anymore. It’s a credibility requirement.

The 11 Claude skills that clean it up

Here’s where it gets practical. This industry pro built a workflow of 11 skills, each one aimed at a specific weakness. What I like is that it’s a full pipeline, not one magic button. Each step does one job well.

1. /writer writes the first draft in your style, not Claude’s default voice.
2. /editor strips out the dead giveaways so every sentence and word earns its place.
3. /fact-checker removes fake stats and invented quotes. Say one hallucinated number out loud in public and nobody trusts you again.
4. /anti-AI style is the default rewrite that breaks the obvious, predictable AI rhythm.
5. /ban-the-AI-words audits the classic offenders like “delve,” “crucial,” and “tapestry.”
6. /ban-the-AI-patterns restructures the go-to phrasing traps like “It’s not X, it’s Y.”
7. /sound like your posts takes your 5 best written posts and learns to write like you.
8. /humanizer catches any AI words and phrasing that slipped through the earlier passes.
9. /red-pen flags the weak lines before a single reader ever does.
10. /self-critique has Claude audit and critique its own output. Claude checking Claude.
11. /auto-block-banned-words is the safety net, so nothing on your red list ever sneaks through again.

How to actually use this

You don’t need all 11 running at once to get value. Here’s how I’d apply what the author laid out:

• Start with the tells. Print the list of 12 and keep it next to you. Do one manual editing pass hunting only for those words.
• Read your draft out loud. The single fastest test. If a phrase feels stiff in your mouth, it reads stiff on the page.
• Feed your own voice in. The “write like you” idea is the real unlock. Grab your 5 strongest posts and use them as the style anchor so the output starts from your rhythm, not a generic one.
• Layer the checks. Draft, then edit, then fact-check, then humanize. Separate passes catch far more than one giant prompt trying to do everything.

I think the deeper lesson from this contributor is bigger than any word list. AI writing isn’t bad because it’s AI. It’s bad because people ship the raw first draft with no edit. The tool gets you 80 percent of the way, and the human closes the last 20 percent that actually makes it sound like a person. That last stretch is where trust lives.

That’s also why this fits where the whole industry is heading. As more feeds fill up with unedited machine output, the writing that sounds genuinely human is going to stand out more, not less. The editing pass is becoming the competitive edge.

The original post has the creator’s full reasoning and the little one-line jabs for each tell, which are worth reading on their own. Go check out the full LinkedIn post for the complete breakdown, then run your next draft through the list before you hit publish.