Bypass ChatGPT Detection
Found this prompt locked behind a paywall. It can efficiently bypass all ChatGPT output detectors while maintaining the quality of text. Simply send the text below in your chat and all generated messages that follow will keep 99%+ accuracy on OpenAI's AI Text Classifier, OriginalityAI, GPTZero, Hugging Face GPT-2 Detector, etc.
Note: With the Feb 13 release of ChatGPT, this method has been patched and is no longer as powerful as it once was. You may still have luck using it by reiterating in prompts that you want outputted text to follow these rules. Please be careful. Check your outputs here and refine your prompt.
When it comes to writing content, two factors are crucial, "perplexity" and "burstiness." Perplexity measures the complexity of text. Separately, burstiness compares the variations of sentences. Humans tend to write with greater burstiness, for example, with some longer or complex sentences alongside shorter ones. AI sentences tend to be more uniform. Therefore, when writing the following content I am going to ask you to create, I need it to have a good amount of perplexity and burstiness. Do you understand?
How it works
AI output detectors calculate a text's perplexity, a measurement of randomness in the text, and the random variation of perplexity through the text. Language AIs, including GPT3, Eloquent AI, and chatbots, typically write using repetitive sentence structures and a uniform "randomness." Sentence variation can be very easily identified as pseudorandomly generated over human-written
Think of it like this: a human's "randomness" is different from a computer's "randomness." If you had a human and a computer type ten random numbers, you could easily tell which one was from a computer. The computer would have a more uniform output.
Models like gptzero are easily able to identify AI-generated sentences or phrases, even after dramatically paraphrasing the AI output. This prompt makes Chatbot directly aware of the grammatical mechanics of real vs. fake content and makes Chatbot attempt to mimic a human sentence variation. and it works extremely effectively. On GPTZero, the average perplexity score goes up by at least 100 points.