Oct. 16, 2026
Last time, I referenced some conflicting research findings regarding whether or not people can identify AI-generated text on sight.
This time, you get to put yourself to the test!
In transparency, my intuition is that people generally can identify AI writing on sight, at least with better than chance accuracy.
But I am a scientist, after all, so…
Let’s do a quick test to see if YOU can recognize AI writing on sight.
Below are two passages, one written by me, and the other written by Claude. We both wrote on the same general topic.
Without skipping ahead to see the answer, try to determine who wrote which passage. (I’m about a million percent confident that you’ll spot the AI tells immediately, but who knows?)
But even if you can correctly spot the AI, the more interesting question to me is, What specifically gives it away? Try to take note of what tells indicate a human versus a large language model…
PASSAGE 1
The best career advice I ever got was a single sentence.
Early in my career, a mentor told me: “Be so useful that people want to tell others about you.”
No gimmicks. No self-promotion. Just solving real problems for the people around me.
It changed how I show up:
→ I stopped asking “How do I get noticed?”
→ I started asking “Who can I help today?”
Opportunities followed, and not because I chased them.
What’s the best career advice you’ve ever received? 👇
PASSAGE 2
Here’s a career principle that I tell everyone: “If you stick to your principles and work hard, eventually things will work out for you.”
I don’t mean this as general life advice. (You might get hit by a meteorite, or you might have a genetic disease, and no amount of hard work can overcome that.)
But in terms of career success, I have found it to be true in my own life.
In the short run, superficial and self-serving tactics can be effective. But in the long run, it’s the genuine, hardworking people who end up doing awesome work in organizations that they love.
If you think you know the answer, then click the solution below to confirm…
Obviously, the first passage was written by Claude, and the second was written by me. (Also, the second passage really does describe a core tenet of my career philosophy.)
I’m actually viscerally disgusted by how Claude fabricates a “true story” about a conversation that never happened. Please, PLEASE never do this.
To me, the obvious AI tells are things like…
- The short, punchy sentences (all of them).
- The prototypical, “No X. No Y. Just Z.” pattern. Again, very short and punchy phrasing.
- The bullet-point key points.
- The try-hard emojis. (I actually expected more of them!)
- The open-ended question finale.
In isolation, NONE of these would necessarily be a tell. But all combined, it’s painfully obvious.
Now, what are the tells of humanity in my writing?
- Probably there are some grammatical mistakes that I didn’t notice.
- I tend to overuse parentheses in the same way that AI overuses em dashes.
- The asides and caveats (e.g., “Well, unless a meteorite hit you”).
- What else?
If you genuinely got it wrong, then please message me explaining your reasoning (seth.osbornLA@gmail.com). I’m convinced the average person can spot this a mile away, but I’m open to being proven wrong!
