Turnitin, a widely used plagiarism detection tool, plays a crucial role in promoting academic integrity and detecting similarities in submitted documents.
On the other hand, ChatGPT is an AI model that generates text resembling human writing. Here, we will try to answer whether Turnitin can detect ChatGPT and how to prevent this?
Can Turnitin or other AI language models identify text generated by ChatGPT? And if so, what measures can be taken to prevent detection?
Quick Facts
- Turnitin is an academic tool that detects similarity rates in submitted documents and encourages originality in writing.
- ChatGPT and other AI models generate text that closely mimics human writing, creating authentic and realistic content.
- The field of Artificial Intelligence is rapidly advancing, with new models and breakthroughs being introduced regularly. This rapid progress in AI makes the task of detecting AI-generated content, like Turnitin, increasingly challenging.
- Turnitin’s detection accuracy is reliable, with a confidence rate of 98%. If Turnitin identifies a text as AI-generated, it is 98% likely to be written by an AI.
- However, Turnitin may not detect every text written by AI. With the advancements in AI, especially with models like GPT-4, AI-generated texts are becoming more sophisticated, making it challenging to discern them as being written by humans. As a result, Turnitin may not identify every text generated by ChatGPT.
Can Turnitin detect ChatGPT?
No, Turnitin can’t always detect content written by ChatGPT. At the current stage, AI-detection tools are not yet advanced enough to consistently detect all text generated by AI language models.
However, this is an ongoing area of development, and algorithms are continuously being created to detect and address AI-generated content. In this article, we will provide more details and insights into this subject.
Note that while Turnitin may not always detect content written by ChatGPT, the technology is evolving, and detection capabilities may improve over time.
OpenAI had released an AI detection tool
On January 31, 2023, approximately one month after the release of ChatGPT, OpenAI announced its own AI-generated text detection tool through a blog post. However, on July 20, 2023, OpenAI revealed that they had removed this model due to its low rate of accuracy.
From this, it’s evident that even OpenAI, the developer of ChatGPT, currently doesn’t possess a model capable of classifying texts generated by this model or identifying them as ChatGPT-generated. Moreover, with the GPT-4 model, the capabilities of such models have further improved.
Of course, over time, things may change, and algorithms can be enhanced. However, this applies to both detection tools and generation tools.
Turnitin announced a new AI detection tool on April, 2023
On April 4, 2023, Turnitin introduced a new tool in a blog post specifically designed to detect writings generated by AI language models like ChatGPT.
Turnitin’s CEO, Chris Caren, expressed admiration for the progress and potential demonstrated by models like GPT-3.5 compared to previous one. To address these advancements, Turnitin accelerated the development of AI writing detection capabilities in response to the needs and concerns of educators.
Acknowledging the rapid development of artificial intelligence, Caren further emphasized, “We will continue to adapt and respond to the next iterations and innovations in AI writing, putting the safety of students and the needs of educators and institutions first.“
If you’d like, there is a videos session below that explains the features of Turnitin’s tool for detecting AI-generated writing. You can watch it.
Plagiarism checker needs to keep up with rapid development
Generative AI tools, including AI chatbots like ChatGPT, are experiencing rapid growth. Within just two months of its launch, ChatGPT (GPT-3.5) reportedly amassed over 30 million users and received approximately five million daily visits.
Currently, Turnitin’s platform reportedly boasts new features that can detect AI writing with 98% confidence and a false-positive rate of less than one percent in controlled lab environments.
However, 98% accuracy does not mean that all AI-written texts are detected at that rate. It indicates that Turnitin’s AI can accurately identify texts it perceives to be written by AI with a 98% correctness.
However, achieving 100% certainty in detecting AI-generated text remains impossible. Muhammad Abdul-Mageed, a professor specializing in natural-language processing and machine learning at the University of British Columbia, explains that AI language models aim to generate human-like, fluent text. As newer AI models become more powerful and produce even more fluent language, existing detection tools quickly become outdated.
GPT-4 exhibits human-level performance on various professional and academic benchmarks, surpassing the capabilities of GPT-3.5, which scored lower in comparison.
Caren, a representative from Turnitin, emphasizes the company’s commitment to adapt and respond to the next iterations and innovations in AI writing. Turnitin will continuously update its plagiarism checker to prioritize student safety, as well as the needs of educators and institutions.
Take a look at this brief demo presented by David Adamson, an AI scientist at Turnitin. In the demo, he guides you through Turnitin AI writing detection capability.
How does Turnitin work?
Turnitin is a plagiarism detection tool that compares a student’s assignment against various sources, including internet documents, previously submitted papers, and subscription repositories. It creates an ‘Originality Report’ that highlights any matching text between the student’s submission and other sources. Both lecturers and students can access this report.
In addition to checking for similarities, Turnitin stores a record of all submitted assignments in a central database. They do this to check future submissions against assignments submitted previously, even in the subsequent years.
Note that while Turnitin retains a copy of submitted assignments, it does not reproduce or disclose them to third parties. The stored copy is only used for comparison purposes. Students maintain ownership of their assignments, and the content is never shared with external parties.
How to prevent Turnitin detect ChatGPT or AI-generated content?
Here are some tips and techniques to help you avoid a high similarity index in Turnitin:
1. Request Other Iterations from ChatGPT
One effective resource is ChatGPT itself. After receiving an initial response, request alternative iterations from ChatGPT to enhance the text’s uniqueness. Ask for a different wording that specifically targets a lower similarity score in Turnitin.
2. Leverage External Paraphrasing Tools
Using external tools like Grammarly can be a valuable strategy to bypass Turnitin and other AI detection systems.
These tools help transform AI-generated content into high-quality writing that closely resembles human composition.
3. Explore Uncommon Topics and Conduct Comprehensive Research
Choosing a unique and less-explored topic can significantly reduce similarity scores. Since there are fewer sources to match, it’s an effective way to avoid detection.
By implementing these techniques, you can reduce the similarity index in Turnitin and ensure that your work stands out as original and unique.