Prompt:
I asked it to write a 2000 word paper on why EV are bad for the environment.
The first result came out as only 10% human generated content.
I went back and asked it to make the writing have more personality, and less robotic sounding.
Boom, it told me it removed a bunch of technical jargon and added more personality. After testing again, 72%.
Finally, I asked it to add back 10% of the removed technical jargon, and add 3 minor grammatical errors (this ended up being the wrong their, and using 2 apostrophes where they weren’t needed).
After testing that on three ai detectors, it passed them all at well above 90% real.
If you prompt it to "increase its perplexity and burstiness" that will have the desired effect.
Some reading about this (not mine): https://medium.com/@TheJasperWhisperer/the-dummy-guide-to-perplexity-and-burstiness-in-ai-generated-content-1b4cb31e5a81
