Can Turnitin Detect ChatGPT?

Turnitin and GPTZero are leading tools for identifying AI-generated content, specifically capable to detect ChatGPT with high accuracy. However, the occasional misidentification of human-written text underlines the system’s imperfections. These tools reflect a growing commitment to academic integrity in the face of advancing AI writing tools.

New Turnitin Feature For Detecting ChatGPT & AI Writing Tools

Turnitin has released a new AI detection tool that can detect ChatGPT content.
Turnitin claims their tool has a 98% success rate.
Mistakes can occur, with human-written text sometimes misidentified as AI-generated.
The development of such tools signifies a serious stance against AI-aided dishonesty in academia.
GPTZero, a Princeton student’s creation, also aims to detect ChatGPT, though reliability varies.
The usage of these tools is increasing, signalling a commitment to originality amidst the rise of advanced AI writing tools.

Table of Contents

Can ChatGPT be Detected by Turnitin?

Yes, Turnitin can detect ChatGPT. Turnitin’s AI detection technology can see the use of AI writing tools, including ChatGPT. Turnitin claims their AI detection platform can identify the use of AI writing tools, including ChatGPT, with 98% confidence.

Turnitin’s new AI detection feature was launched in April 2023, allowing schools and universities to detect ChatGPT in students’ assignments.

The advancements in AI writing technology have created a sense of unease among educators, who fear that students may use AI writing tools such as ChatGPT to plagiarize their work. However, as AI writing tools develop rapidly, so does the technology behind Turnitin, the leading platform used by universities and educators to check students’ work.

Limitations of Turnitn’s Ability to Detect ChatGPT

There are some limitations to Turnitin’s ability to detect ChatGPT. Turnitin’s detector for ChatGPT faces technical limitations and may not be completely accurate.

Some users have reported false positives, mainly when ChatGPT was only used to write part of the assignment and a human wrote the rest. Sometimes, essays written by mixed AI and human sources were misidentified. Turnitin missed individual sentences written by ChatGPT and falsely identified parts written by a human.

Nonetheless, Turnitin is continuously developing its technology to detect emerging forms of misconduct, including AI-assisted writing and AI writing generated by tools such as ChatGPT.

How Does Turnitin Detect ChatGPT in Assignments?

Currently, Turnitin’s AI writing detection model is trained to detect content from the GPT-3 and GPT-3.5 language models, which includes ChatGPT.

Turnitin’s AI writing detection feature works by focusing on accuracy. When it flags a piece of writing as AI-generated, the system is highly confident in its assessment. The company has worked to ensure a high accuracy rate, aiming for a less than 1% false positive rate to prevent students from being wrongly accused of misconduct.

A false positive in this context refers to the system incorrectly identifying a text written by a human as being AI-generated. While Turnitin strives to minimize these, it acknowledges that there is still a small risk of false positives occurring. In such cases, it’s up to the instructor to apply their professional judgment, knowledge of their students, and the specific context surrounding the assignment to make an informed decision.

The technology identifies the use of AI writing tools, showing how many sentences in a written submission may have been artificially generated. The AI writing detection model may not always be accurate, so it should not be used as the basis for adverse action.

When a Similarity Report is opened, the AI writing indicator is in the side panel. After a short period of processing, the hand will display one of three potential states: blue with a percentage between 0 and 100, yellow with a message indicating that the submission is still processing, or red with a message indicating that the request has failed to process.

Can Universities Use GPTZero to Detect ChatGPT in Assignments?

Yes, universities can use GPTZero to detect ChatGPT in assignments. However, GPTZero is an app built by a computer science student at Princeton, Edward Tian, that can quickly and efficiently see whether an essay is ChatGPT or human written.

The app uses a series of complex and deep algorithms to analyze the text and determine its origin. GPTZero is a ChatGPT detection tool that can be used to find out if something you’re reading was generated with AI or not.

However, it’s important to note that GPTZero shouldn’t be relied on for making official judgements. Errors are highly likely when GPTZero detects AI or ChatGPT in documents. Work needs to be checked manually by someone who understands how ChatGPT’s watermarking works.

Accuracy of GPTZero in detecting ChatGPT

The accuracy of GPTZero in detecting AI plagiarism is still being assessed, but anecdotally, it seems to be working well for most people. The developer, Edward Tian, never claimed that GPTZero is perfect, and it’s essential to remember that errors are possible.

While the tool has demonstrated success in detecting AI-generated text in some cases, such as identifying a LinkedIn post as bot-written. Users should still check the work for mistakes and use their judgment when relying on GPTZero’s results.

How GPTZero Detects ChatGPT in Student Assignments

GPTZero determines the difference between human and AI-generated text by analyzing two indicators: perplexity and burstiness.

Perplexity measures the complexity of the text; if GPTZero is perplexed by the text, it has high complexity and is more likely to be human-written. Burstiness refers to the uniformity of randomness within the text. By analyzing these properties, GPTZero can identify whether the reader will likely be written by a human or an AI like ChatGPT.

How to Use ChatGPT and Pass Turnitin AI Detection

Using ChatGPT to complete schoolwork automatically is considered cheating, and Turnitin has developed technology to detect AI-assisted writing and AI writing generated by tools such as ChatGPT

However, there are ways to bypass AI detection, such as making the text more emotional and adding misspelled words to make it look more like human writing. Here are some tips to avoid AI detection and pass Turnitin AI detection:

Use ChatGPT to create a rough draft or an outline for your work.
Change words, sentences, and text to your writing style to make it look legitimate.
Use Grammarly and apply spelling and grammar corrections to your document. Grammarly rewards sections of your article, changing them enough to reduce the AI detection score. Grammarly is one of the quickest ways to help avoid the detection of ChatGPT.
Avoid using ChatGPT to generate the entire essay. Make multiple edits and change the punctuation. This will cover up ChatGPTs watermarking.
Use a website like ZeroGPT to check your essay before submitting it to Turnitin.
It is important to note that cheating is not ethical and can have serious consequences.
If you are using ChatGPT to write code for university assignments, you should restructure the code, swapping the order of functions.

What is Turnitin’s AI Content Detection feature?

Turnitin’s AI Content Detection feature is a tool designed to help educators maintain academic integrity by detecting if a piece of writing has been generated by an AI, such as ChatGPT. When a paper is submitted to Turnitin, the submission is broken into segments of text that are run against an AI detection model. This model gives each sentence a score between 0 and 1 to determine whether it is written by a human or AI. The model then generates an overall prediction of how much text in the submission it believes has been generated by AI.

How does Turnitin’s AI Content Detection feature work?

The AI Content Detection feature works by breaking a submitted paper into segments of text, which are then run against Turnitin’s AI detection model. This model assigns each sentence a score between 0 and 1, with 0 indicating that a sentence was not generated by AI and 1 indicating that it was entirely generated by AI. The model then uses the average scores of all the segments within the document to generate an overall prediction of how much text in the submission has been generated by AI.

What parameters or flags does Turnitin’s model take into account when detecting AI writing?

Turnitin’s model is trained to detect differences in word probability between human writing and AI writing. AI models like GPT-3 and ChatGPT generate sequences of words based on picking the next highly probable words, resulting in a consistent and highly probable fashion. Human writing, on the other hand, tends to be inconsistent and idiosyncratic, resulting in a low probability of picking the next word in the sequence. Turnitin’s classifiers are trained to detect these differences.

Can I check past submitted assignments for AI writing?

Yes, previously submitted assignments can be checked for AI writing detection if they’re re-submitted to Turnitin. However, only assignments that are submitted after the launch of the AI writing detection feature (4th April 2023) are automatically checked for AI writing detection.

What languages are supported by Turnitin’s AI Content Detection feature?

Currently, Turnitin’s AI Content Detection feature only supports documents submitted in long-form English. If a non-English paper is submitted, the detector will not process the submission.

Will I need to pay more for this AI Content Detection feature?

The first iteration of the AI writing detection indicator and report are available to exist Turnitin customers as part of their existing licenses. However, beginning January 1, 2024, only customers licensing Originality or Turnitin Feedback Studio with Originality will have access to the full AI writing detection experience.

What is Turnitin's AI Content Detection feature?

Turnitin's AI Content Detection feature is a tool designed to help educators maintain academic integrity by detecting if a piece of writing has been generated by an AI, such as ChatGPT. When a paper is submitted to Turnitin, the submission is broken into segments of text that are run against an AI detection model. This model gives each sentence a score between 0 and 1 to determine whether it is written by a human or AI. The model then generates an overall prediction of how much text in the submission it believes has been generated by AI.

How does Turnitin's AI Content Detection feature work?

The AI Content Detection feature works by breaking a submitted paper into segments of text, which are then run against Turnitin's AI detection model. This model assigns each sentence a score between 0 and 1, with 0 indicating that a sentence was not generated by AI and 1 indicating that it was entirely generated by AI. The model then uses the average scores of all the segments within the document to generate an overall prediction of how much text in the submission has been generated by AI.

What parameters or flags does Turnitin's model take into account when detecting AI writing?

Turnitin's model is trained to detect differences in word probability between human writing and AI writing. AI models like GPT-3 and ChatGPT generate sequences of words based on picking the next highly probable words, resulting in a consistent and highly probable fashion. Human writing, on the other hand, tends to be inconsistent and idiosyncratic, resulting in a low probability of picking the next word in the sequence. Turnitin's classifiers are trained to detect these differences.

Can I check past submitted assignments for AI writing?

Yes, previously submitted assignments can be checked for AI writing detection if they're re-submitted to Turnitin. However, only assignments that are submitted after the launch of the AI writing detection feature (4th April 2023) are automatically checked for AI writing detection.

What languages are supported by Turnitin's AI Content Detection feature?

Currently, Turnitin's AI Content Detection feature only supports documents submitted in long-form English. If a non-English paper is submitted, the detector will not process the submission.