One of the most common questions I get is, “How can I check to see if something has been written by AI?”
I put this to the test yesterday to show how an “AI humanizer” tool takes AI-generated copy and rewrites it (using AI, of course) to trick AI detection tools.
TL;DR: It’s Pretty Darn Easy to Make AI Content Look Human-Written
The first generic drafts of my posts were caught by all the AI detectors I tried. But when I asked ChatGPT or Bard to apply “the style of Beth Z, Your Nerdy Best Friend” or if I just asked the tool to “Make this sound more human,” I was able to fool several of the AI detector sites. And running them through an “AI humanizer” tool helped me fool even more. The more I ran it through stuff and asked for tweaks, the more “human” the text became, and that was without me creating any content myself.
Watch this video to see the whole test.
Trying to “Catch” People Using AI Fosters an Environment of Distrust
A person in one of my sessions yesterday asked, “We are looking over submissions from our essay contest. Some of them are definitely suspicious. Should we disqualify them?”
Watch my response here: