Depending on which AI checker you look at, I either passed with flying colours or failed miserably.
I was curious, so I ran a test
I’ve been thinking a lot about AI in recent months, while I’ve been hard at work on the manuscript for my next book. Word by word, line by line, I’m making progress every day. Slower than a GPT can write, but (in my humble opinion!) with vastly superior quality.
Out of curiosity, I ran 1500 words of my latest writing through a few AI detection tools.
The results were fascinating.
My text was rated by four tools as:
- 100% human-generated (which it absolutely was)
- 85% AI-generation score (definitely not)
- 75% signs of AI-generation (nope)
- 24.4% possible AI-generated text (nah, bro)
The 85% AI score stopped me in my tracks. The reasons given? The text was structured, consistent, error free, polished, and the given example progressed logically through the stages described in the text (of course it did – I made it up, so I could make it “behave” any way I wanted it to!)
The tool’s analysis finished with this gem: “The text maintains the same level of polish and formality throughout its considerable length, which is unusual for human-written content of this scope.”
Got it. So being able to write a couple of decent paragraphs is now a sure-fire sign of AI generation. Spare me!
Are you relying on AI tools?
If you’re relying on AI tools to tell you whether something is AI-generated, consider what the model is looking for. AI produces structured, consistent, polished output because that’s how good writers write, and that’s what it has been trained on. I tried multiple different passages on that 85% tool, and it gave me wildly different scores, even though all of the passages were 100% written by me.
Long story short: don’t take one tool’s analysis as gospel, especially if you’re going to assess/judge/discipline people for it. We need to use our critical thinking skills, now more than ever!
