ChatGPT is an AI tool that lets you create texts on almost any topic. ChatGPT works by “learning” from a huge amount of data and then creates a new text.
The output text that you get is usually not plagiarized (directly copied from another source) so there is a new need for a tool that can detect if a text is original or not. This type of tool is called an AI Detector.
This is necessary to stop people from passing off ChatGPT’s work as their own (Think about students using ChatGPT to do their homework).
Let’s find out more about AI detectors.
What Are AI Detectors?
An AI detector is a tool developed to distinguish AI-written text from human-written text. They often also use artificial intelligence to spot linguistic patterns.
Our tool Winston AI uses both linguistic analysis and data training to test whether a piece of content was written by AI or not. Read more about how our tool works here.
We have a 99.98% accuracy rate at detecting AI content. You do need to be careful as many other tools available do not have these accuracy rates and are therefore not very useful.
Do AI detectors detect ChatGPT?
Yes, AI detectors can detect text that is produced by ChatGPT.
Let me show you how it works:
First I will ask ChatGPT to write me an essay.
Next, let me test this text with our AI Detector Winston AI
As you can see, our AI detector correctly predicted that this text was written by an AI tool.
There are in fact many AI tools on the market, not just ChatGPT. There are tools like Claude which is developed by Anthropic, Bard, which is developed by Google and LLaMA which is developed by Facebook.
It is important to note that AI detectors give results based on probability. So sometimes, you will get a result like “Human Score 35%”. This does not mean that 35% of the text was written by a human. It means that the probability of the text being written by a human is 35%.
If you are interested in knowing which parts of the texts our AI detector thinks were human-written and which parts of the text were AI-written, you can use the AI prediction map.
The red text predicts AI text and the green predicts Human text.
Does Turnintin detect ChatGPT?
Turnitin is an educational company that has specialized in software that detects plagiarism.
Turnitin launched its AI detector in early 2023. They claim to have an accuracy rate of 98% but there have been many unhappy customers who have shown examples of when Turnitin incorrectly flagged students’ work as AI.
Vanderbilt(A University from Nashville,) has decided to stop using Turnitin and many students on Reddit have been falsely accused of cheating because of the tool.
I turned it in and waited a week to get my grade back and to my utter shock, I got a ZERO. I read the assignment feedback which said my cover letter was “100% AI written” according to TurnItIn. WTF??!!!! I wrote the ENTIRE thing myself so to say I was shocked was an understatement.
https://www.reddit.com/r/college/comments/17x954b/i_was_falsely_accused_of_ai_cheating_and/
Turnitin can detect ChatGPT but there are also a significant amount of cases where it detects human-written content as AI. For that reason, I recommend using Winston AI.
Can you copy and paste from ChatGPT without plagiarizing?
Yes, you can copy and paste from ChatGPT without appearing as plagiarism on a plagiarism tool. The text that ChatGPT outputs is different from any other text.
Plagiarism is when you directly copy from another text whereas ChatGPT is not copying from other texts but using machine learning to predict the next word and producing a text based on your prompt.
This does not mean you can use ChatGPT to do your work for you as an AI detector like Winston AI can show that the text was written by AI.
Can you bypass an AI detector?
While there are many tools out there attempting to bypass AI detectors, none of them are very good. Some of them are called “Humanizers” or have something like “Undetectable AI “ as their selling point.
Don’t waste your time with them as they still appear as AI content on our detector Winston AI and also make the text significantly worse quality.
You can probably bypass some AI detectors that have low accuracy rates but those AI detectors are not worth using anyway as they don’t work.
Frequently Asked Questions
Yes, Turnitin’s AI detector can probably identify text generated by ChatGPT and other AI writing tools with high accuracy. Turnitin launched this AI detection tool in 2023. Their AI analysis will flag papers as “highly probable AI generated text”.
Turnitin’s AI detector analyzes writing patterns such as vocabulary, syntax, and coherence to determine if text was AI-generated. Turnitin has trained its AI on over a billion student papers to recognize human vs AI writing styles.
It’s unclear if Turnitin will retroactively scan old submissions. Some sources suggest past papers may be flagged if resubmitted. To be safe, don’t reuse old AI-generated text.
Turnitin doesn’t disclose their exact AI matching thresholds. However, reports indicate it needs a very close match to flag AI content, perhaps 70%+ overlap. Unique paraphrasing helps avoid detection.
While paraphrasing may help, Turnitin’s AI analysis goes beyond verbatim text matching. Their advanced detection can identify AI writing style even if well paraphrased. Use caution.
