Can Turnitin Detect ChatGPT?

Have you ever found yourself gazing at a blank screen and having no clue as to where to start your assignments? The deadline is looming, and you start to feel the heat. We have all been in that position. Sometimes, with the pressure of deadlines, students will turn to AI tools like ChatGPT to help them draft their assignments. ChatGPT is quick and clever, and it can produce any text in seconds, but there is just one question many students are often concerned about: can Turnitin detect ChatGPT?

In this blog, we will tell you if ChatGPT is detectable and how Turnitin detects ChatGPT. We'll also discuss the accuracy of Turnitin, the limitations of AI detection, what this means for students, and when it can act as an assignment helper for brainstorming ideas. By the end, you will have what you need to understand how to use AI responsibly and the risks of being flagged by detection tools. So read till the end!

What is Turnitin, and how does it work?

For a majority of your academic career, you have probably sent dozens of essays and papers through Turnitin, but what is Turnitin exactly? 

Turnitin is a global leader in academic integrity, and its primary function has been to help educators check student work for plagiarism.  When you submit a document in Turnitin, it reviews the text and checks it against a very large database of documents (academic papers, journals, websites, and previously submitted student assignments). After the checking process, it generates a Similarity Report, which indicates what text matched an existing source and provides a percentage score to indicate how closely your work resembles other works (the higher the number, the more your work resembles other sources). 

The Similarity Report is a very useful tool for educators to help identify potential plagiarism, but it is also useful for you to use to check your work prior to submission. 

The question of whether Turnitin can detect ChatGPT became a concern with the introduction of AI tools. In light of AI, Turnitin updated its platform to include AI writing detection. It can not only check for plagiarism, but it can also evaluate the text to determine if it was likely generated by an AI tool.

Can Turnitin Detect ChatGPT Content?

Yes, Turnitin can detect ChatGPT content to a certain extent. The company started adding its AI detection function in 2023, where it uses advanced algorithms to identify writing patterns similar to machine-generated writing.

When you enquire, 'Is ChatGPT detectable?', the answer is 'Yes, but not 100%'. The system uses patterns in unusual phrasing, tone, and structure. ChatGPT also consistently produces overly balanced sentences and often continuously uses many coordinating conjunctions: "Moreover", "Additionally", "Therefore".

Furthermore, Turnitin provides a confidence score that you can colour code:

1. 'Red' means 'AI-generated', highly likely

2. Yellow means there is mixed writing with AI and human writing

3. 'Blue' means that it is most likely human writing

Interestingly, Turnitin can even pick up differences across ChatGPT variant models (naming GPT-3, GPT-3.5, and GPT-4), but all accuracy varies.

How Does Turnitin Detect ChatGPT?

You may be curious: how does Turnitin catch ChatGPT, exactly? The primary mechanisms include: 

Pattern Recognition: Turnitin’s AI detector looks for strange patterns in sentences, unvaried word choice, and phrases indicative of using AIs. Human writing frequently has imperfections, while AIs make grammatically correct but often too generic phrases. 

Database Matching: The student's work is put through hundreds of billions of online sources, academic papers, and prior students’ essays, and is checked for any similarities. 

Stylometric analysis: The tool can compare the “style fingerprint” to see if the tone, structure, and transitions an author shows typical human inclinations or if it seems to be mimetic of AI outputs. 

Probability Score: Instead of an actual yes or no, Turnitin also provides a number from 0 to 100%, which the professors can use to determine if a specific section was AI-created. 

Multi-layer detection: By encompassing both plagiarism detection and AI content detection, Turnitin adds both flagged copied text and flagged AI-written text to make it harder for students to circumvent the detection system.

The Accuracy and Limitations of Turnitin’s Detection

While the answer to whether Turnitin can detect ChatGPT is yes, no tool is perfect. You may be thinking to yourself, 'Does this mean that ChatGPT is detectable every time?' The answer is that there are some things you should know. 

False positives: AI is sometimes (and often is) not detected because it is falsely flagged as AI by humans. This generally happens more often with very formal or short writing. Although Turnitin boasts low false positives (~1%), human reviewers still have to risk being as tightly regimented to determine if something is human or AI.

The 300-word minimum: Turnitin's AI detection algorithm works best with prose of 300 words or more. Turnitin's inabilities increase with lower quantities of word lengths in assignment lengths. Short answers, summaries, or offering non-prose (bullet points, tables) format also complicate the matter further. This is a noteworthy detail relevant to the overall question of how Turnitin detects ChatGPT.

Human Editing and Paraphrasing: When students edit and rewrite AI-generated text in their style, the likelihood of detection is dramatically decreased. Their voice and ideas complicate the detection. 

Evolving AI Models: New models, like GPT-4, mean detection tools must continually evolve. This may mean that new AI outputs can evade detection as they are beginning to be produced, and, therefore, asking if ChatGPT is detectable becomes a continually moving target.

What This Means for Students

For UK university students, the fact that ChatGPT is detectable should serve as a wake-up call. Rather than learning to use AI as a shortcut, here's how to use it safely:

Think of AI as an aide rather than a substitute: Use ChatGPT for brainstorming, organising ideas, or generating some practice questions for essay topics. If you submit a whole essay written by ChatGPT into your submission, that's plagiarism.

Edit in your voice:  Always take AI-generated content and review and rewrite it so it reflects your distinct voice and level of comprehension. Doing this will help you to avoid getting flagged by Turnitin. 

Include your examples and analysis:  AI struggles when it comes to providing personal examples from your lived experience. Including your examples makes it more likely to avoid detection and enhances your academic writing. 

Understand academic integrity:  Universities in the UK take plagiarism and misuse of AI technology very seriously. If you submit assignments written entirely (or significantly) by AI and do not acknowledge this practice, you could be penalised. 

Ask for expert support when necessary: Rather than depending on artificial intelligence, engage with a professional assignment help in the UK, such as Locus Assignments. We provide personal support that will help you learn and produce original pieces of high-quality work to submit.

Conclusion

So, can Turnitin detect ChatGPT? Yes, although there are plenty of limitations. Turnitin can identify AI-generated text, but it is not always accurate, and, at times, it may detect genuine work as AI-generated. For students, this means carefully and responsibly using AI tools. Instead of relying on technology completely, rely on your own critical thinking, writing skills, or use access to AI technology when it's needed as a helper for a project to brainstorm and put your project together. The better you are aware of the advantages and disadvantages of AI detection, the better you can ensure your originality and avoid potential unnecessary risks.