Most AI writing isn't bad. It's just obviously not human. That's the problem nob
By Maria Gharib · October 17, 2026 · Curated by George's Blog
Most AI writing isn't bad.
It's just obviously not human.
That's the problem nobody talks about clearly enough.
It's not that Claude writes badly.
It's that it writes too cleanly.
Same sentence rhythm.
Same openers.
Same patterns every single time.
And readers spot it now.
Detectors spot it even faster.
I've been building in AI for a while now.
And the thing that keeps showing up,
Whether I'm reviewing content for our team or reading what other founders produce,
Is this the case: the writing is technically fine, but it feels hollow?
You can't put your finger on it immediately.
But something is off.
Turns out there are 29 specific patterns that give it away.
I went through all of them.
The infographic covers every single one.
Here are the ones that matter most if you want your AI writing to actually pass:
→ Significance inflation.
Words like "pivotal" and "groundbreaking" everywhere.
Just say what happened.
→ Promotional adjectives.
"Vibrant", "seamless", "breathtaking".
Say what you actually mean instead.
→ The AI word list.
"Delve", "landscape", "crucial", "showcase."
These are dead giveaways.
Replace them.
→ Em dash overuse.
That reflex to connect everything with a dash.
Use a comma or a full stop.
→ Sycophancy.
"Great question!" as an opener.
Remove it entirely.
Just answer.
→ Filler phrases.
"Worth noting" and "important to remember."
Cut them.
Start after the filler.
→ AI sign-offs.
"Hope this helps."
"Let me know",
"Let's explore."
End with something real or nothing at all.
→ Generic conclusions.
"The future looks bright."
State your actual conclusion in one plain sentence.
There are 21 more in the infographic.
The faster fix is setting up Claude properly before you ever generate a draft.
Custom styles, user preferences, a Claude Project with your voice baked in.
One prompt before you generate anything changes how much cleanup you're doing afterwards.
Honestly, most people skip the setup and then wonder why the output sounds wrong.
The infographic walks through all of it.
The 29 tell how to fix each one and the full system for keeping AI patterns out consistently.
Learn AI for free: https://lnkd.in/exrZ97Q3
Worth a look if you're producing a lot of written content with AI right now.
If this is useful,
You'll probably like what I post the rest of the week.
Send this to someone on your team who's using AI to write and keeps getting caught out by the output.
Curious which of the 29 hits closest to home for you.