Asking how you would go about detecting OpenAI’s ChatGPT is essentially asking how to detected any AI writing. ChatGPT is the model AI model, the standard algorithm every AI detector is programmed to to understand.What exactly is that programming though? How do AI Detection tools point to a piece of text and say this is AI-generated content or this is human-written content. What AI Detectors do is measure certain stylistic trends in a piece of text. We’ll discuss which trends and how in this article.
And now, all we have to do is take the ChatGPT output and input it into AI Detect and click Detect to get a result:
As you can see, AI Detect determined the text was 1% Human and given it was so clearly AI writing, the readability score of 38% also shows us how robotic the writing style feels.
Now we’ve increased our human score to 86% and readability score to 65%.
Table of Contents
- What is ChatGPT?
- What is AI Writing?
- What is an AI Detector?
- How to Detect ChatGPT
- How to make AI-written text Undetectable
- Conclusion
- FAQ
What is ChatGPT?
OpenAI’s revolutionary large language model changed the world forever, and with the capability of generative AI technology like it to create articles and essays in an instant, it was inevitable that students and professionals would start using AI generated text for their work.What is AI Writing?
There is an AI writing style just like there is a human writing style. The machine learning of artificial intelligence chatbots like ChatGPT comes from both the vast swath of writing available online as well as the huge datasets being input into ChatGPT everyday.ChatGPT specifically is programmed to write with simplicity and precision in its word choice, sentence structure, and sentence length to make every output easy to understand. These writing elements are conducive to a robotic feel and low readability. Humans write more naturally, with complexity and randomness.What is an AI Detector?
What AI checkers do is measure for the perplexity and burstiness in a piece of text and if those metrics are uncannily low, it’s probable from an AI writing tool. Perplexity and burstiness are the watermarks of writing from LLMs. The problem is though, these metrics can be present in human writing as well. This is why AI detectors all have varying rates of false positives that often make accusations of AI plagiarism easy to dismiss. Turnitin, for example, has a 4% false positive rate on the sentence level.It begs the question, do AI content detectors even work? The answer isn’t a clear cut yes. Numerous universities have stopped using AI detectors and most students have moved on from using free AI services like ChatGPT which get detected, in favor of undetectable AI services that allow them to generate text that bypasses any detector tools.How to Detect ChatGPT
It’s rather easy. You can probably be your own ChatGPT detector by finding word choice that is too precise or simple. Still, most people will just employ an AI content detection systems to give them a result that backs up their suspicion.One of the best AI Detectors is AI Detect, which allows you to access a free tool at first, with just as much power, if not more, than Turnitin. So, lets generate an essay with ChatGPT to run through AI Detect. Here it is:How to Make AI-Written Text Undetectable
If you remove the watermarks of AI writing and replace them with the watermarks of human-writing, such as complexity, randomness, and naturalness, you can make AI text undetectable. The process of this is called humanizing. This is why AI Detect also has a humanizer.Since most students and professionals need content that’s human-like enough to not face consequences, undetectable text is the most optimal text for use. Here’s what our ChatGPT output looks like if we humanize it with AI Detect. Simply click “Humanize Text” and you’ll get this: