Loading...
How to spot it
The signs of AI writing in ad copy.
Nine tells give a robot away in an ad. Here is what each one looks like, how often it showed up when four leading AI models wrote 400 ads for us, and how to fix it.
By Ad Legends ·
The surprise
The famous AI words are mostly gone.
The words everyone jokes about showed up in only 6% of the ads we collected, and not one ad shouted in capitals. Today’s models have learned to avoid the obvious tells.
The habits they kept are quieter, and far more common: the long dash in 52% of ads, a headline that is mostly the brief in 66%, and one of the same five buttons in 62%. That is where to look now.
The nine tells
What to look for.
Most common first. Each share is the part of our study’s ads that showed the tell.
The big one
Is the em dash a sign of AI?
Often, yes. 52% of the ads in our study used the long dash, more than any other tell. A single dash proves nothing. A dash in most sentences is the most reliable tell left.
It is also the easiest to fix. When we added a ban list to the ask, dash use fell from 50% to 0%.
Detection
Can you tell if an ad was written by AI?
Not reliably, and you do not need to. Tools that guess who wrote a text struggle on thirty words, and a person can write slop as easily as a robot can write a clean line.
Look for the tells instead. They are what your customer notices, whoever wrote the ad.
Now what
Count them, then fix them.
Paste your ad into the free slop meter to count the tells. Use the tested ask to write a cleaner draft. Then hand it to a Legend for the part a meter cannot count: the idea.
Questions
A few more questions
What are the signs of AI writing in an ad?
The long dash in most sentences, a headline that repeats the brief, a button anyone could write, an emoji nobody asked for, exclamation marks, big claims with no proof, and an opening that invents your problem. In our study of 400 AI-written ads, the dash alone showed up in 52%.
Is the em dash a sign of AI?
Often. It was the most common tell in our study, in 52% of AI-written ads. One dash proves nothing; a dash in most sentences is a strong sign.
Do AI models still use words like the famous ones everyone jokes about?
Rarely, in ads. Those stock phrases showed up in only 6% of the ads in our study. The tells that remain are the dash, the repeated brief and the same few buttons.
Can AI detectors tell if an ad was written by AI?
Not reliably on short copy like an ad. Check the tells instead; they are what readers notice.
How do I remove the signs of AI from my ad?
Change the ask, then edit. A tested ask with a ban list cut ads with a tell from 71% to 2% in our study. Then fix what is left by hand and check it on the free slop meter.
