Asking ChatGPT To Put Perplexity And Burstiness In An Essay Appears To Fool AI Detectors

Funny how the solution to fool detectors is to tell the AI to write less like an AI.

Check all AI results for accuracy, fairness, and potential harm. Ultimately, it’s human oversight that safeguards responsible use.

After discovering that there's a way to effectively cheat AI detectors (from watching this video by Success With AI) I went about and did my own tests on it.

These are the results.

First, I gave it my initial prompt:

Burstiness & Perplexity

Then I proceeded to teach it what role burstiness and perplexity play in human writing. Then it was time to tell it to rewrite a text I'll give it, while bearing those in mind.

Re-Writing

After copy-pasting its previous story, it proceeded to re-write it into this:

  • In my opinion, no human writes like this anyway, so it's not very good.

AI Detectors + Original Text

https://gptzero.me/ "likely entirely by AI" 

AI Detectors + Improved Text

https://gptzero.me/ "likely entirely by human" 

.

Although detectors are rapidly advancing and trying to catch up to AI, it is worth noting that there is now publicly available content that teaches people how to trick them.

I believe this will be a never-ending game between AI Generators and AI Detectors of one-upping each other.