How Does Turnitin Detect AI?

The market for AI technologies is vast, amounting to around 200 billion U.S. dollars in 2023. It is also expected to grow to over 1.8 trillion U.S. dollars by 2030, based on the Statista reports. This is because of the growing demand for AI, which is also used for writing.

With the increasing use of AI in writing, many educators are worried that the students are using the tools for their projects, affecting their submission originality.

That is why Turnitin is here to help. However, how does Turnitin detect AI and plagiarism?

Educators want students to learn, understand, and think critically instead of using AI tools that undermine the educational process.

This blog will include everything you need to know about Turnitin, how it helps detect AI, and how it works.

What is Turnitin?

Turnitin has served the global education community since 1998 and launched AI writing detection capabilities in 2023.

It is an AI detector tool that helps educators identify the originality of students' work that they may have done using generative AI models.

Their work is compared with a wide database of academic content, publications, and web pages. This platform offers the best solution to manage academic integrity.

Many teachers and educators use this tool to check and detect whether their students have used any AI tools, such as ChatGPT, for their assignments and projects.

Understanding AI in Writing

New tools and technologies are continuously introduced to make the work easier for each individual, as are the tools available for writing.

AI writing tools, like ChatGPT, Jasper, Gemini, and others, help generate human-like text. They are popular among students because these tools help them produce essays, reports, and creative writing pieces.

However, it leads to their lack of understanding and thinking as they become entirely dependent on such tools, which can later not benefit them because assignments and projects are assigned to students to learn something new and use their creativity and understanding.

How Does Turnitin Check for AI?

For 25 years, Turnitin has been increasing its reach to help educators, publishers, and researchers identify whether AI writing tools like ChatGPT or text spinners have been used in students' submissions or assignments.

It used advanced technologies to detect the content and determine whether it was AI-generated. Tunritin’s AI content checker helps ensure the student's submissions are original.

The Turnitin helps show the overall percentage of AI writing tools used in the document and highlights the text segments written using AI tools.

Many educators use it for all their writing assignments, and the tool shows copied text and sources. Many AI business ideas also use such AI detection tools. If its blue box shows that AI is 0%, this means that the assignment doesn’t contain AI-generated content.

However, it provides data for educators to decide based on their academic and institutional policies. Turnitin also says not to use its percentage of the AI writing as the only basis for action or grading measures.

Natural Language Processing (NLP) Techniques

Turnitin uses NLP to analyse the structure and style of the text, which helps distinguish human-written content from machine-generated.

Machine Learning Algorithms

Machine learning helps Turnitin improve its detection capabilities. The algorithms are trained on several datasets, allowing the system to learn the commonly used words or patterns in AI writing. Their training helps them recognise such patterns.

How Does Turnitin AI Detection Work?

When a student's paper or assignment is submitted to Turnitin, it is broken into text segments of around a few hundred words.

These segments allow us to capture each sentence in the submission by overlapping the segments with each other.

Turnitin then scores each sentence between 0 and 1 to check if it is human-written or AI-written. It will receive a score of 1 if it is generated by AI and a score of 0 if it determines that a sentence is not generated by AI.

After a score of all the segments has been analysed by the tool, the overall text in the submission is checked, and the AI score is generated, which the tools believed the submission included.

Turnitin AI detecting tools are trained to detect content from ChatGPT's GPT-3 and GPT-3.5 language models.

Example

» Imagine a student submits an essay on “The Marketing Trends of Digital Marketing” and submits it on Turnitin. The Turnitin AI checker will start to compare it with several documents, web pages, and papers related to digital marketing trends in its database.

» Its AI detection algorithms will kick in and analyse the writing structure and flow.

» It will check whether the content has a natural flow, personal opinions, and some small grammatical mistakes or is formal and includes perfect grammar, robotic sentences, and patterns or words used in AI text.

» It will divide every sentence into segments and then thoroughly check the overall text. It will show the percentage of content that is human-written or AI-generated.

What Does the Different Colour Indicator Mean?

You can open a Similarity Report that you can see on the side panel of the AI indicator. After the content or text is processed, the AI writing indicator will show the state of your content, whether it is AI-generated or human-written.

» Blue with a percentage of 0 or between 20 and 100: It shows the submission has been processed, and the percentage shows all the texts generated by AI. However, this doesn’t mean it is the percentage of the whole submission.

This AI tool claims that percentages between 1 and 20 may sometimes show false interpretations. That is why it displays an asterisk (*) for this percentage to tell the score is less reliable.

» Gray with no percentage displayed (- -): If the grey box is displayed with no percentages, the Turnitin AI detection tool indicates that they can’t process the submission, leading to you not having any access to the AI writing report.

It can be because the user could not meet the file requirements needed to submit the submission for AI writing detection. If it shows the error, it means that Turnitin could not process your submission successfully.

Turnitin AI Writing Report & Submission Breakdown

Turnitin will generate an AI writing report that includes the overall percentage detected as AI. This will indicate as the qualifying text in the document submitted by the user that the AI writing detection model of Turnitin will identify it as AI-generated or could be generated by AI and could be paraphrased using AI tools.

The Submission Breakdown

The text detected as AI is divided into two detection categories, but first, let’s learn about the qualifying text. The qualifying text is the single sentence in paragraphs that make up a longer written work, such as an essay, article, etc. Here are its two categories.

» AI-generated only- The qualifying text generated from a Large Language Model (LLM) is detected in this category. The text will be highlighted in a cyan colour in the submission breakdown bar.

» AI-generated text that was AI-paraphrased- This category will indicate that the qualifying text was AI-paraphrased, meaning it was modified using an AI-paraphrasing tool, such as Quillbot. The text of this category is marked in purple colour in the submission breakdown bar and the submitted content.

Can Turnitin Detect Quillbot?

QuillBot is a popular paraphrasing tool that rephrases sentences and enhances the user's writing by providing synonyms or alternative sentence structure.

Many students make use of such tools to avoid detection by plagiarism software. That is why the important question arises: Can Turnitin identify text altered or generated by QuillBot?

QuillBot helps rephrase text, improves vocabulary, and offers different sentence variations so the users can choose the one that suits their content better. It analyses the text and rewrites suggestions while maintaining the original sentence's meaning by altering its structure and wording.

However, whether Turnitin can detect the altered text by QuillBot depends on various factors:

» Extent of Alteration: It may not detect the AI if your text is altered significantly through QuillBot, such as by changing words or rephrasing sentences. But if the sentence ideas stay the same, it may flag your work as similar to the original source.

» Quality of Paraphrasing: QuillBot's paraphrasing effectiveness can help tell how Turnitin can detect similarities. Turnitin can detect AI text and flag these sections if your sentence is similar to the original content or common phrases used in writing.

» AI Detection Capabilities: Turnitin can detect and identify patterns in writing that help them flag the paraphrased content with some characteristics of the AI-generated text.

Indicators of AI-Generated Text

Some factors indicate the content is AI-generated and Turnitin also looks for while detecting the text or submission as follows:-

» Lack of Personal Voice

Ai-generated text will often lack the personal touch that comes with human writing. It will feel robotic and include words that are sometimes hard to understand.

Turnitin will show the content as human-written if the students express their views, experiences, and emotions while doing their project or assignment.

» Repetitive Structures

When we use AI tools, we notice that they often include repetitive phrases or sentence structures. They rely on predefined patterns, making the expression of content less diverse.

» Overly Formal Langauge

AI-generated writing may involve a formal tone that doesn’t match the student's writing. The language used is polished or has technical words not required for the assignment, indicating to the educators that the student’s submission has AI-generated text.

Turnitin’s Technology Explained

Now, we know the workings of the Tunitin and how it is the best AI text detector tool, especially for professors and educators. Even one of the faculty members at Johns Hopkins University uses this tool for all his writing assignments. It is helpful because certain technologies and algorithms are used to analyse the submitted work.

1. Database of Known Sources

» With the help of its vast database that contains many academic papers, journal articles, and web pages, Turnitin can compare the student's submissions with these known sources.

» It can happen because Turnitin trains its model on the data spread over time, including AI-generated and authentic academic writing. This academic writing is taken from across geographies and subject areas.

» The model also has a database of second-language learners, English users who don’t speak English, college students, universities with wide enrollments, and less common subject areas like sociology, anthropology, geology, etc. It helps Trintin to lower bias while training the model.

2. Algorithmic Comparisons

Turnitin’s algorithms analyse the text against its database and model, which look for matches or similarities after the submission.

It then calculates the percentage and shows the score that indicates whether the submission and how much of it matches the existing content. It is an original AI that helps check the authenticity of the work.

Turnitin VS Other Detection Tools

Turnitin is one of the many AI checker tools available online, but its features make it different from others.

Tools like Grammarly and Copyscape help check the plagiarism and grammar of the content, while this AI text detector involves more thorough analyses using the database and algorithms.

It allows educators to manage academic integrity and assignment originality by providing the best AI detector platform.

Its data enable it to compare student assignments and projects to ensure accuracy and reliability.

Limitations of Turnitin’s Detection

While Turnitin is a beneficial and best AI detector tool, it still has some limitations. Here are some of its limitations.

False Positives

The Trunitin shows accurate data, but its results are believed to be 98% accurate, meaning there can be false negatives and positives.

Original work may sometimes be flagged as AI-generated if it includes elements of machine writing. False positive occurs when fully human-written is identified as AI-generated.

However, the risk of false positives is less than 1%, with over 20% likely AI-generated content for the document.

It can be a limitation, but Turnitin take measures to avoid such false positives. No score or highlights are assigned for AI detection scores in the 1% to 19% range.

Language Limitation

Turnitin only supports English and Spanish languages. So, it can detect those documents submitted in long-form English or Spanish.

Spanish documents have also been submitted since September 2024, and till then, they were only available for English submissions.

If non-English or non-Spanish documents are submitted, the detector will not process them and will indicate an error state.

That is why if you want to submit it in any other language, you can’t do so because Turnitin will generate no report if the submitted content does not include English or Spanish language.

Conclusion

Today, everyone, whether a writer or a business, is using AI tools because they make their work easier. The students also use AI writing tools that lower the value of the submission.

It is because when the students do the assignments, they learn new information and gain knowledge. So, what can the educator do? They can also use tools that detect AI in their submission, such as Turnitin.

But how does Turnitin detect AI? It helps educators identify the originality of students' work that they may have done using a generative AI tool. It is the best AI text detector tool for professors and educators.

It is helpful because certain technologies and algorithms are used to analyse the submitted work. So, Turnitin's workings can help maintain the originality of student submissions and help educators check the authenticity of submissions.

Frequently Asked Questions

Q1 How does Turnitin know if you used AI?

Ans. Turnitin knows if you used AI because it uses a statistical model that helps analyse submitted papers. It helps determine if the paper was AI-generated and detects content from the ChatGPT.

Q2 How accurate is Turnitin’s AI detection?

Ans. Turnitin’s AI detection is accurate but not foolproof. It uses various techniques to assess content originality. It detects the content while keeping the false positive rate under 1% for documents. They might show a human-written text as AI-written for one out of every 100 fully human-written documents.

Q3 What percentage of AI is acceptable on Turnitin?

Ans. If Turnitin shows a score of 20% or higher, it means your document is AI-written. It also no longer displays scores below 20%.

Q4 If students use Grammarly for grammar checks, does Turnitin detect it and flag it as AI?

Ans. No, Turnitin doesn’t target Grammarly-generated spelling, grammar, and punctuation modifications to content. It shows AI content written by LLMs such as GPT-3.5.