12 AI Writing Giveaways and 12 Claude Skills That Fix Them

This morning I scrolled LinkedIn and saw three posts in a row that “delved” into something. Three. That’s when I knew I had to share this one.

A LinkedIn creator just posted a sharp list of the 12 most obvious signs that AI wrote something. They also shared the 12 Claude skills they use to clean up text that sounds too much like AI. The author’s complaint is simple. Their whole feed is AI slop, and nobody bothers to edit it. I think they’re right, and the list is painfully accurate.

The post comes in two halves. First come the words to delete. Then come the tools that catch them for you. I’ll go through both and add a few tips on how to use them together.

🚫 The 12 AI Giveaways to Delete Today

The original poster’s request is direct: delete these from your writing today. Here’s each one with the author’s take and a bit of context on why it gives you away.

1. “Delve.” Nobody talks like that. When did a coworker last say they’d delve into the quarterly numbers? Swap it for “look at” or “dig into.”
2. “Crucial / pivotal.” Almost nothing is that pivotal. If every point is crucial, none of them stand out. Save the big words for the one thing that really earns them.
3. “Tapestry.” The author admits they didn’t even know this word existed. It’s what you get when a model reaches for something poetic and plain words would work better.
4. The fake-suspense opener. You know the one. It’s the “here’s the…” line that promises a big reveal. The creator’s reply: what thing? Say it. Lead with the point.
5. “Hope this helps.” It doesn’t. If your content helps, readers will notice without the reminder.
6. “After careful consideration.” Please do not. It sounds like a rejection letter and tells the reader nothing.
7. “To provide a quick update.” Just give it. Announcing the update only wastes space before the update itself.
8. “Most people…” The original poster calls this straight-up lazy. It’s a vague setup that pretends to know what everyone thinks.
9. “Robust / seamless / realm.” This is the corporate AI starter pack. The words sound impressive and say almost nothing.
10. Adverb abuse. Think of lines like “X quietly runs Y.” That “quietly” adds drama the sentence never needed.
11. “It’s not X, it’s Y.” The famous AI sentence. Once you notice this contrast pattern, you’ll see it everywhere.
12. The dramatic reframe. The author’s example is “This isn’t a budget. It’s a statement of intent.” Their verdict is one word: No.

The expert’s bigger point is easy to miss. You don’t have to use AI to write like you. You can just write yourself. The skills below fix text that AI has overwritten. They don’t replace your own thinking.

🛠️ The 12 Claude Skills That Clean It Up

Some quick background if Claude Skills are new to you. A skill is a saved set of instructions you add to Claude, and you run it with a slash command like “/humanizer.” Each skill does one job. The 12 the author uses cover the whole writing process: planning, drafting, editing and final checks.

1. /copywriting: Writes headlines, copy and CTAs using copywriting rules. It’s a good place to start when you need words that persuade.
2. /linkedin-post: One skill that drafts, edits and then scores your LinkedIn post before it goes live. The author’s standard is to write as if every word pays $20 to earn its place.
3. /fact-checker: Checks for fake stats and made-up quotes, so you don’t have to do it all yourself. This one matters because AI loves inventing a convincing number.
4. /delete-ai-words: Removes the obvious AI words like “delve,” “crucial” and “tapestry.” It’s basically the giveaway list above, automated.
5. /hooks: Rewrites your first two lines. The creator’s warning: you really don’t want to open with that fake-suspense line.
6. /personal-voice: Adds the writing style you’ve fed it, so the output sounds like you and not like a generic assistant.
7. /my-viral-post: Goes through your LinkedIn post analytics and writes like your past outliers, the posts that did far better than the rest.
8. /humanizer: Strips out any AI words that slipped through. Think of it as a final sweep.
9. /red-pen: Highlights the weak lines before you send a draft to someone.
10. /grill-me: Asks you critical questions before it writes a word. This is my favorite idea on the list, because most AI slop starts with having nothing specific to say.
11. /viral-recipe: Works out WHY a LinkedIn post went viral, then reuses that recipe for your own post.
12. /voice-dna: You upload a batch of your writing samples, and it builds a profile of your voice.

⚙️ How I’d Stack These Together

Twelve skills can feel like a lot. Here’s the order I’d try, based on how the author describes each one.

1. Build your voice profile once with /voice-dna and /personal-voice.
2. Start each new post with /grill-me so you’re clear on your actual point.
3. Draft with /linkedin-post or /copywriting, then sharpen the opening with /hooks.
4. Run /delete-ai-words, then /humanizer to catch the leftovers.
5. Finish with /fact-checker and /red-pen before you hit publish.

The study-type skills, /my-viral-post and /viral-recipe, work best as a monthly habit. Use them to learn what your audience responds to, and don’t run them on every single draft.

💡 Prompt of the Day

No skills installed yet? You can still use the author’s checklist. Paste your draft into any AI chat with this prompt of mine:

“Review this draft and flag every AI giveaway: words like delve, crucial, pivotal, tapestry, robust, seamless, realm; fake-suspense openers; ‘It’s not X, it’s Y’ contrasts; dramatic reframes; filler like ‘hope this helps’ or ‘to provide a quick update’; vague ‘most people’ claims; and unnecessary adverbs. Suggest a plain rewrite for each one.”

Personally, I think this post works because it’s honest. The tools help, but the real fix is caring enough to edit. Go check out the full LinkedIn post and see how many of these 12 giveaways are in your last few posts.