Why Your AI-Written Posts Sound Like AI (and 9 Tells to Remove Today)
Nine tells.
Photo by Tara Winstead on Pexels
AI-written posts sound like AI because they share the same few habits: a tidy slogan at the end, a stock "nobody tells you" opener, em dashes, a question on every reply, and made-up personal stories. Remove those nine and most of the smell goes. Below is each one with a bad and a better version. You can paste any draft into the free checker to see which ones it has.
What are the 9 tells?
| # | Tell | Bad | Better |
|---|---|---|---|
| 1 | Tagline ending | "Consistency isn't a habit. It's an identity." | "I posted every day for a month. Day 3 and day 19 were the worst." |
| 2 | "X is the part nobody tells you" | "Burnout is the part nobody tells you about." | "Nobody warned me I'd dread opening the app by week three." |
| 3 | "It's not X, it's Y" | "It's not about followers, it's about trust." | "Followers are nice. The two people who messaged me are why I kept going." |
| 4 | Em dashes everywhere | "I tried it — it worked — twice." | "I tried it. It worked twice." |
| 5 | A question at the end of every reply | "Love this! What's your take?" | End when you're done. |
| 6 | Flattery opener | "Great point, this really resonates." | Start with your actual point. |
| 7 | Round, confident numbers | "This will 10x your reach." | "My reach about doubled. One account, six weeks." |
| 8 | Copying the post you reply to | Restating their sentence back to them. | Add something they didn't say. |
| 9 | Invented stories | "Last week a client told me..." (no client) | Only say what happened to you. |
The ninth matters most. A made-up story is the fastest way to lose someone's trust once they notice.
Why do these tells get noticed?
Readers can't always name the pattern, but they feel it: everything is too tidy, too certain, too smooth. Real people trail off, hedge and say small specific things. A reply that sounds like a bot gets scrolled past or, worse, gets you a reply saying "this is AI". A real example of one getting flagged is in I got called out for sounding like AI.
How do I check my own post?
Paste it into the free "Is this AI?" checker. It gives a human score out of 100, plain-English flags for what it found, and highlights the sentences that read as AI. It's free, needs no login, and takes about ten seconds. For replies specifically, start with how to reply so people see it.
Humanish runs this kind of check on every draft before you see it. Drafts that fail are thrown out, not patched. Nothing is posted for you.
How does Humanish help with this?
It writes in your own voice, learned from your past posts, and checks each draft before you see it. In our own blind tests, judges scored the drafts 3.2 to 3.5 out of 5 for "sounds human", and the share with invented details came down from 27% to 14%. That is better, not perfect. Funny, sarcastic and warm tones are still best effort.
Check a post now · Start my free trial
FAQ
Can AI detectors tell if my post is AI?
Tools like ours give a score, not a verdict. What readers notice is the style, and these nine tells are the style.
Is it bad to use AI to write posts at all?
No. It is bad to post something that doesn't sound like you or says things that didn't happen.
Can I make AI writing undetectable?
We don't promise that, and nobody honest does. The aim is writing you'd be comfortable having said.
Are em dashes always a tell?
One is fine. Three in a short post is a tell.
What should I do first?
Delete the last line of your next draft. If it was a slogan, the post got better.
Try it on your own account
Free for 7 days, no card needed. You send every reply yourself.