#Bypass Any AI Detector
1 messages ยท Page 1 of 1 (latest)
From my experience, this just makes the text corny and doesnt make it sound more human at all. I've made prompts that humanize text before and here are the main steps:
- use synonyms and change word choice a lot
- change sentence structure majorly
- determine what parts of the text sound weird or unnatural and change the synonyms use to more natural sounding ones
- change the main points of the text (if it applies to the text it is outputting)
I created that prompt a long time ago when oai still used the gpt4 4k context model. Back then when GPT4 was much better, it could get 100% on originalityai v1.4 consistently. Now, since GPT4 performance was degraded significantly, it can only get around 20% but still gets 100% on v2 (v1.4 is much stricter than v2 and i would recommend you test your text on v1.4 before doing anything with it). I feel like if i updated my prompt it could still achieve 100% human on v1.4. What I'm trying to say is that its still possible but you have to use a more sophisticated prompt. Also use originalityai to test, its currently the best AI detector rn.
The problem is copyleaks
Every school teacher uses it and it's really good
But yours works to it just takes more tries
did you read what I said? use originalityai
Yeah but teachers in my country use copyleaks! That's the problem! But originality ai works with yours to
We both did a nice job!
Turn yours into a GPT!
no it doesnt, it got 100% ai on originalityai
eh, i would have to update it first and i cba rn
Yeah I said that it works with originality ai!
Ok
???
i just said it doesnt work
Wait I will try it! I mean ZeroGPT and Copyleaks works, but I will test originality to!
I've used your prompt and copyleaks detected it as AI Content!
But it really surprised me, because it sounded human!
๐คทโโ๏ธ
Mine works now with the new update!
I think if you would make it a GPT it would be more accurate
really? did you update your gpt?
Yeah, I tested it three times. It had a bug that I noticed now ๐
100% on 1.4 and 99 on 2
good job!
this basically just tells the gpt "AIs are serious and robotic. make your text more weird and expressive". this "bypasses" detectors easily but it has trouble writing in a formal tone
when you tell gpt4 to act like that it is very obviously ai
It doesn't work against Japanese because one Japanese paragraph will be treated as one word lol
it has multi language
Ah. I needed to login, right?
I just tried out the demo lol
yeah you get 50 free credits iirc
thanks, it works
However, the result was weird and yet interesting.
Here's what i got when tested my personal GPTs speaking Japanese, and tested my own input
So I guess I am a "GPT" - a version of... lol
so your own input was flagged as ai and output from your gpt was flagged as human?
weird