Abstract

With the development and availability of AI models there is concern about AI plagiarism in higher education. To help address this challenge, AI detectors such as Turnitin have been created to help faculty assess the presence of AI plagiarism in various assignments. This study aims to present faculty-student communication strategies that in combination with flexible use of Turnitin may help address AI plagiarism in higher education. To this end, this study involves an exploratory content analysis of a sample of 149 cases of AI-related plagiarism from 35 undergraduate social science course sections, with 901 enrolled students across three universities in the Southwest. Three instructors identified cases occurring in April 2023 through October 2025. All of these cases operated under the syllabus stipulation that AI should not be used to directly generate assignment text. Instructors used the Turnitin AI detector to help identify potential cases where the student used AI to compose major portions of a written assignment. Out of 149 cases of AI-plagiarism, only one case involved denial of misuse of AI. The 149 cases of AI-plagiarism were generated by 142 individual students. Students flagged generally did not repeat AI plagiarism in the classes taken with the instructors involved in this study. We present a basic statistical summary of characteristics of the cases identified. In addition, the content analysis identified five recurring themes reflective of various communication strategies that seem helpful when coupled with the flexible use of the Turnitin AI detector. Such strategies are illustrated via actual examples of instructor-student discussions. We conclude that utilizing the Turnitin AI detector, along with certain communication strategies, might be an effective way to detect and reduce the frequency of AI plagiarism in courses.

Introduction 

Assignments carefully planned by faculty to incorporate AI usage may be helpful for student learning. For instance: “a professor may present students with AI-composed text and ask them to assess the AI work along several criteria”; or “an assignment may have students work with AI to generate a response perhaps through a chain or series of prompts and then assess how well the AI is able to compose the text according to criteria <specified by the instructor>” (Canisius University, 2024). These are simple examples of an assignment using AI that cuts across many areas including social sciences. Many faculty (perhaps most in social science areas) do not incorporate AI in their assignments and use traditional non-AI based essays, papers, and/or discussion forums in their courses. In both an AI-based and non-AI based assignment where the student is required to compose their own response to the questions posed, AI plagiarism is possible.

With the emergence of Large Language Models and various AI tools, there is increased concern about AI plagiarism in many universities and colleges across the United States and elsewhere. As noted by Oravec (Oravec, 2023, p.213), “AI has added new dimensions to academic cheating challenges as students can easily access powerful systems for generating content that can be presented in assignments, exams, or published papers as their own.” Based on their literature review, Perkins, et al (Perkins et al., 2024, p. 93) note that “the advancement of AI has sparked serious ethical concerns related to the potential misrepresentation of authorship and authenticity of academic work”. They also state that “such actions undermine the core principles of academic integrity, erode the educational value of assignments, and compromise academic institutions credibility.” (Perkins et al., 2024, p. 93).

On the one hand, the use of AI to perform various tasks in the workplace may mean that students need to learn how to manage AI and assess its products. On the other hand, many believe that allowing students to let AI write their assignments may undermine the development of critical thinking skills and understanding of concepts within various disciplines. As the Chair of the Sociology Department at the University of California at Berkeley indicates: “Some students now rely on AI to summarize readings or draft essays. That work is an important part of the learning process, and they won’t learn if AI does it for them.” (Ma & Ballardo, 2025).

AI detectors (both commercial and publicly available) can be used to assess the presence of AI plagiarism. They vary in overall accuracy, the degree of false positives, and in the degree of false negatives. All three dimensions are of interest. However, the extent of false positives is probably of special interest in the higher education setting. No detector is totally foolproof.  However, in two studies Turnitin had no false positives given the methodologies employed (Walters, 2023; Weber-Wulff, 2023 ). Walters found no false negatives by Turnitin while Weber-Wulff found a false negative rate of 27.8%. However, in Weber-Wulff, Turnitin had the second best false negative rate (that is, the undetected rate for AI generated papers) among the 14 AI detectors tested. This may reflect Turnitin's concern about avoiding false positives as much as is reasonably possible.

This paper has three overall research questions:

  1. What are basic characteristics of the case sample (e.g., AI percentages, extent upper level vs. lower-level undergraduate courses)
  2. To what degree can the Turnitin AI detector be fairly and flexibly used to address AI plagiarism?
  3. What are the major recurring themes with respect to constructive instructor-student communication strategies used in conjunction with Turnitin results? What are examples of actual instructor- student dialogues that illustrate these themes?

Research Question 1 is addressed in a section on descriptive statistics. This section provides basic background data such as number of course sections involved in this study and average enrollment per course section.

Research Question 2 is addressed in a couple of ways. Certain figures in the descriptive statistics section provide some of the evidence suggestive of the impact of using a variety of communication strategies along with Turnitin on reducing the extent of AI plagiarism. For instance, this section shows how rarely students flagged for AI plagiarism deny AI misuse. Also, the Introduction cites two recent studies that support the reasonableness of using Turnitin to help identify AI plagiarism cases. Examples of actual instructor-student dialogues help reinforce the statistical findings on effectiveness.

Research Question 3 is addressed in part by a content analysis of 149 case sample of AI plagiarism cases. This analysis helps identify major themes in communication and related Turnitin AI detector use that seem to be helpful in addressing these cases. Communication strategies are illustrated with items such as example syllabus statements as well as actual instructor-student email dialogues.

We present cases to illustrate our themes. While the themes are recurring, the cases are meant to appear in the sections presenting the dialogue between professors and students sometimes even between students themselves. We are using these examples to highlight potential strategies. We acknowledge that these strategies are developmental, and we expect professors to adapt and experiment with them in their own classes.

Overview of the Turnitin AI Detection Process

The Turnitin AI detector (once activated by the instructor) will produce a report for each individual assignment submitted. The report will show a similarity percentage and an AI percentage. The similarity percentage for an assignment reflects the extent of copying text from published materials without attribution. The AI percentage reflects the extent of use of one or more AI tools to compose assignment text or to escape detection. For a particular assignment, the report also provides color shading of sentences and paragraphs that reflect AI use. The individual report may be shared with the particular student.

The AI detector works on assignments but does not work for online discussion forums.  For discussion forums (including the main post), the instructor must look for one or more telltale signs (e.g., robotic phrasing, excessive rambling, formulaic bullet points, vague or overly general writing, writing that is unusually sophisticated compared to other writing submitted by the particular student). The instructor in such instances can create a “dummy” essay assignment box, copy and paste the suspicious piece into the box, and then run the AI report. This is given additional coverage in the Theme 4 portion of the findings and discussion section.

Fundamental steps in the communication process

Syllabus Statement: To establish clear expectations, the syllabus should include a section on AI plagiarism. This should define AI-generated writing, explicitly prohibit its use for specific assignments, provide examples of restricted tools (e.g., ChatGPT, Quillbot), and outline penalties for violations. An example statement is provided in Theme 1 of the findings and discussion section.

First Week Reinforcement: Instructors should reiterate AI-use policies in an announcement during the first week of class. This serves to reinforce the syllabus content and ensure all students understand the rules. Theme 2 includes a model announcement for reference.

Setting Enforcement Thresholds: Instructors need to decide what AI percentage will trigger intervention. Turnitin does not display percentages below 20% to reduce false positives. Based on analysis of 149 documented AI plagiarism cases, the instructors begin addressing issues at 30%, typically issuing a warning and reducing the assignment grade. Halbert et al., (2025) state that “since AI is so new, we use the 30% threshold to give us some wiggle room for error.) (p.7).

At 50% or above, instructors usually assign a failing grade for the assignment and issue a formal warning. While most students stop after a first offense, repeat violations may lead to an F in the course and referral to the Dean of Students for further investigation and disciplinary action. Additional details are discussed in Theme 4.

Methodology

We employed a qualitative content analysis approach to examine AI-related plagiarism, aiming to enhance understanding and contribute uniquely to academic discussions surrounding the rise of AI-generated writing since November 2022 (Bryant & Charmaz, 2007; Corbin & Strauss, 2015). Our methodology focused on analyzing real-world plagiarism cases and uncovering themes that could guide educational policy and instructional strategies.

Data were collected through instructor responses to open-ended questions such as: “Can you briefly describe the impact of early communication with your students during the first week of class?” and “Can you briefly describe any cases involving tools like Grammarly, unverified academic sources, or other relevant issues?” Responses and documentation were analyzed using a three-step coding process: open coding, axial coding, and thematic development (Aicinena & Ziyanak, 2021; Soyer & Ziyanak, 2018).

This study reviewed 149 AI-related cases flagged by the Turnitin AI detector, with IRB approval obtained from the University of Texas Permian Basin. Three social science instructors from different universities in the Southwest participated, each submitting flagged assignments spanning from April 2023 through October 2025. Among the 149 cases, 81 included email communication between the instructor and student, with the student generally acknowledging the misuse of AI. The remaining 68 cases involved only instructor communication, with no student response nor denial of the flagged AI misuse.

To ensure confidentiality, pseudonyms were used for both instructors and students, and institutional identities were not disclosed. The cases were grouped based on similarities, leading to the emergence of five main themes: “Plagiarism policies and syllabus reminders”, “announcements for early interventions”, “unverified sources”, “the need for reliable AI detection tools”, “clear, firm, and fair communication”.

These themes are presented in the findings and discussion section, where selected conversations between instructors and students are quoted to highlight key issues and teaching strategies. We employed a coding process to analyze email conversations and instructor reflections (Strauss & Corbin, 1998). All relevant documents including emails, instructor notes, and written experiences were organized into a master file (Elliott, 2018; Soyer & Ziyanak, 2018). During open coding, we identified incidents of AI plagiarism, triggers for detection, whether students admitted to violations, cheating tactics used, and AI detection percentages.

Next, in axial coding, we refined and grouped emerging ideas to align with our research questions. We identified 43 initial codes, such as: cheating, AI tools, matching percentage, confession, apology, ChatGPT, Turnitin, and academic dishonesty. These were consolidated into 12 axial codes, focusing on student behavior, instructor intervention, use of detection tools, and patterns in communication.

In the final thematic development stage, we reviewed and connected these codes, solidifying five central themes (listed above). This process provided a robust framework for understanding the challenges instructors face and the instructional strategies that help promote academic integrity in the age of AI.

Findings and discussions

We categorized our findings from 149 AI-related cases into five categories. In this section, five themes will be illustrated. These cases will be exemplified through direct quotes from both instructors and students to demonstrate how instructors applied strategies and to clarify how AI was used in the students' work.

Descriptive Statistics 

  1. The case sample includes 149 AI Plagiarism cases generated during the study period (April 2023 through October 2025).
  2. These 149 cases represent about a 17% AI plagiarism rate measured as the number of cases (149) divided by the number of student enrollments (901) (where 1 student enrolled in 1 course section is 1 student enrollment)
  3. These 149 cases were generated by 138 students (non-repeaters) who did not have additional AI plagiarism cases in other courses taught by the instructors during the study period; there were only 4 “repeaters” who generated 11 AI plagiarism cases during the study period. Essentially, very few students who were flagged and communicated with using the strategies presented in this paper continued with further AI misuse.
  4. There were 35 undergraduate social science course sections offered by the three instructors during the study period.
  5. Ten course sections were face to face and 25 course sections were online. Eight course sections were lower level and 27 were upper level.
  6. The average course section enrollment level was about 25.7 per course section (901 student enrollments divided by 35 course sections)
  7. The average AI plagiarism percentage among the 149 AI plagiarism cases was 87%.
  8. The average similarity index score among these 149 cases was 20%.

Theme 1: Plagiarism Policies and Syllabus Reminders

This theme revolves around having a syllabus that presents clear plagiarism policies and has explicit description of the relatively new area of AI plagiarism. This is necessary so that in the event of AI plagiarism the faculty member can refer to these policies in communications with the student. Such a statement needs to incorporate a listing of example tools that can be misused for AI plagiarism purposes. That is for computer generation of assignment text or for disguising such misuse (e.g. Quillbot, Grammarly, ChatGPT, and so forth). The listing also needs to be updated as new forms of AI emerge. Excerpts from a plagiarism statement used by one of the instructors follows below:

All forms of academic dishonesty (including AI Plagiarism) will be addressed in accordance with institutional policies and procedures...Any suspected violation will be investigated and possibly reported to the Dean of Students.

AI Plagiarism involves the use of AI tools to literally generate assignment text and/ or to disguise such use.  Examples of AI tools include packages such as ChatGPT, Canva, Synthesia, ChatSpot, Auris AI, Jasper, Notion, Bard, Grammarly, Grammarly GO, and so forth.

Students found responsible may receive a failing grade ("F") for the course.  At a minimum, the extent of misuse may affect the level of penalty in a given assignment.

While the syllabi used by all three instructors indicated the right to give a student an F for the course in the case of plagiarism including AI plagiarism, this was not often invoked.  Instead, instructors either assigned an F or "0" to the assignment, or applied a grade reduction after discovering the extent of the issue.

Dr. L sent the following to Student E:

Dear Student E:

I have thoroughly reviewed your assignment report and based in part on the Turnitin AI detector report it appears that the content was 100% AI generated.  For now, I am withholding your AI report from the Dean of Students. I would like to hear your response regarding this matter.  You now have the AI score.  Please refer to the course syllabus for detailed guidelines on this issue.  (Note: If the student does not respond they receive a reduced grade or possibly an F for the assignment depending on the extent of misuse and whether or not the incident is the first instance of such behavior).

I am offering you a significant opportunity to learn and improve.  With effort and commitment, you can still achieve a decent grade 
and successfully complete the course. I hope you understand the seriousness of this situation and take the necessary steps to address it in future assignments.

Best Regards.

Dr. L

When students raised questions about what is in the syllabus, the instructors would refer them back to specific, relevant portions of the syllabus. Dr. L went further by referring students to the syllabus quiz which included questions about plagiarism and the special case of AI plagiarism. This quiz is mandatory for Dr. L's students in order for them to access the course modules.

Theme2: Announcement for Early Intervention

One of the instructors utilized the announcement tool to provide a fair warning to students during the first week of class. This was done immediately after quickly grading their papers and preparing feedback for each student. The announcement served as a proactive measure to address any concerns and clarify expectations early on.

Dear Students,
As we begin the course, I want to address an important concern about this week’s assignment. While I know some of you from previous classes, for others, this may be your first time with me. After reviewing submissions, I noticed signs of AI-generated content, which is troubling. As stated in the syllabus and grading rubric, using AI to complete assignments is strictly prohibited. My stance is not against AI itself, but rather against plagiarism.

Dr. A.

Another instructor didn’t wait for the first assignment. He addressed it on the very first day. He went over the syllabus thoroughly essentially repeating it but with his own take. He made sure to do this right away on the first day of class. The instructor essentially just copied the material. He did, however, provide and explain what AI plagiarism is and emphasized that it cannot be excused. He also provided some examples and mentioned other related tools or devices.

Theme 3: Unverified Sources 

Students need to be informed that AI generated references used to help address an assignment need to be treated with caution. The existence of such references needs to be verified. Sometimes AI-generated references are for articles or other documents that do not exist. Beyond this, students need to know that such references may not back up what AI is stating in its response to prompts and thus they need to read the actual document if in fact it exists.

The first case in this section involves AI plagiarism discovered in a main post in a discussion forum. As presented in more detail in the overview of the Turnitin AI process, the instructor saw telltale signs of possible AI plagiarism and thus used a special (though unwieldy) process to check for AI plagiarism necessary for discussion questions and not for essay assignments. In this case, 35% of the main post was shown by Turnitin to be AI-generated. The student had directly copied sentences and page numbers from certain alleged references into the main post. The instructor found these references were fake.

Dear R,

Thank you for your post.

I was unable to verify your sources and the information you highlighted.  Can you send me the DOIs for the articles cited?

Here are some of the sources you cite:

Kupfer, Marcy. 2017. "Cultural Significance of Fast-Food Advertising." Sociological Quarterly 58(4): 567-589.

Valtorta, R., et al. (2023). Gender Identity in Advertising: A Critical Analysis. Journal of Gender Studies, 32(4), 123-145.

Advertising plays a significant role in shaping and reinforcing gender identities (Valtorta et al. 2023).

The simplicity of the jingle underscores the idea that a straightforward, catchy tune can have a lasting impact on consumers (Kupfer 2017).

Dr. L.

Upon receiving the letter above, the student admitted to the irresponsible use of AI in this initial discussion forum.

In this case also involving student R in a discussion forum, Dr. L reports several students indicated via email that they suspected student R may have engaged in AI plagiarism (e.g., they saw telltale signs such as robotic writing) and was citing unverified sources that do not exist (e.g., a certain article in Sociological Quarterly 58(4): 567-589). Dr. L contacted Student R due to the concerns of these other students in the course. Dr.  L also emailed the three students thanking them for indicating their concerns and that follow-up communication with Student R would take place promptly. In fact, in one email, one of the concerned students sent the following interesting note to Dr. L:

Dr. L

Thank you for taking my correspondence seriously.  I respect and value my education and want to engage with other students who feel the same way.  The only reason I noticed that the source was unavailable was because I wanted to read it and couldn't find it!

Student P

Theme 4- The need of reliable AI detectors 

Evidence of AI detector accuracy and effectiveness is provided in the studies by Weber-Wulff et al. (2023) and Walters (2023) as cited in the introduction. This is further supported by the finding that of the 149 cases in the study identified as having significant AI use (typically greater than 30%, which is above Turnitin’s recommended 20% threshold), only one case continued to deny that the text presented as their own work was generated by AI. Turnitin now employs a threshold of 20% for the possibility of AI use in submissions, a policy implemented starting Fall 2024. When presented with the Turnitin AI output showing what was AI generated in the piece reviewed, all others acknowledged significant misuse of AI.

Turnitin (2025) indicates that it emphasizes specificity to the extent that a few potential AI overuse cases are missed. Turnitin places a high value on preventing false positives as much as possible; its testing shows that there is only a 1% false positive rate among those receiving at least a 20% AI usage score. To be additionally conservative, instructors providing their cases usually used a 30% AI usage score before imposing any penalties.

Theme 5: Clear, firm, and fair instructor communication

Theme 5 is touched upon in the discussions presented in the other themes. However, it may be helpful to reinforce this as a distinct theme.

Clear, firm, and fair communication includes:

  1. Promptly notifying the student of results of the Turnitin AI detector when it provides evidence of AI generation in an assignment (i.e., AI percentage greater than 30% with delineation of telltale signs if they are present).
  2. Stating the AI percentage and either providing a copy of the AI Report that shows the plagiarized passages or summarizing where the bulk of the apparent issue resides (e.g., the last 40 lines of a 60-line essay).
  3. When the occasional student response challenging the findings occurs, referring the student to the relevant portions of the syllabus including the various forms of what constitutes AI misuse (e.g., students sometimes do not realize that Grammarly or in general AI detectors count as AI misuse); this may also involve citing the results of the Walters and the Weber-Wolff studies reinforcing the validity of using the AI Turnitin detector to assess AI plagiarism presence.
  4. Assessing factors such as whether or not the incident is a first time or repeated occurrence, and the extent of AI misuse; based on such information students are either given a reduced grade for the given assignment or in extreme cases an F for the given assignment (Note: students generally do not get a D or F in our courses from only one instance of AI misuse; to get a D or F in a course requires repeat AI misuse and/or poor performance in other components of the course grade (e.g., Exams).
  5. First incidents are not reported to the Dean of Students for investigation (students are instead told that the Dean of Students for now will not be informed about the first incident of AI plagiarism).

As noted elsewhere in this paper, the result has been that very few students in our case sample are repeaters of AI plagiarism in the courses taken with the instructors in this study during the study period. Beyond this aspect, some students express remorse or accept the results but note the special circumstances leading to AI misuse. For example, some excerpts from a student response to an instructor’s initial communication are below.

Dr. G:

I sincerely apologize for the high level of AI detected in my writing.  I appreciate your decision not to report me this time.  I was completely stuck on how to navigate this paper.  I recognize this does not excuse the outcome of this matter. I am committed to ensuring that my future work is entirely my own with adequate citations as appropriate.

Student U

Conclusion

AI-based assignments when properly planned and structured by faculty can be a helpful part of the learning process. We provide two general examples of this in the introduction of this paper. Students may understandably use AI as an assistive resource just like an encyclopedia or relevant book.  However, to facilitate the learning process, faculty usually require students to synthesize relevant materials and generate their own text with appropriate citations. AI plagiarism occurs when a student uses AI to generate the text for major portions of a written assignment and/or uses AI to disguise this use.

As shown in the descriptive statistics section:

  1. Among those enrolled in the 35 social science course sections in this study, the overall AI plagiarism rate was 17%.
  2. Among the142 students who were flagged for one or more incidents or cases of AI plagiarism there was only 1 student who continued to deny AI misuse even after correspondence with them. In addition, only 4 students were “repeaters”. 138 students flagged did not repeat their misuse of AI during the study period in the various courses taken with the study instructors. This data preliminarily suggest the potential effectiveness of using an AI detector such as the Turnitin detector along with various communication strategies to reduce the extent over time of AI plagiarism.

            Communication strategies are presented and illustrated throughout most of this article.  Some examples include:

  1. Having a syllabus that establishes clear expectations regarding academic dishonestly in all forms with additional information about AI plagiarism which is a relatively new form of academic dishonesty.
  2. Instructor use of a reliable AI detector such as Turnitin along with familiarity with telltale signs to help assess instances of AI plagiarism; this includes among other things prompt notification of students when misuse occurs and description of the extent of AI detected and where in the assignment it has primarily occurred.
  3. For a first incident, indicating to the student that “for now” this will not be reported to the Dean of Students. instead, depending on the extent of misuse, there will be some markdown of the grade for the given assignment.
  4. Other strategies as presented and illustrated throughout most of this paper (e.g., in the sections on the five themes identified through content analysis, and so forth).

This study provides preliminary evidence of effectiveness of using an AI detector along with certain communication strategies. However, it has some limitations that need to be addressed in future research efforts. A key limitation is that it involves only three social science instructors at three universities in the Southwest. To improve generalizability, this study needs to be replicated with a larger number of universities and faculty across additional disciplines.

In the introduction, we noted the findings of two recent studies that tested a variety of AI detectors. However, AI packages seem to be rapidly changing and may provide increased capabilities that may increase student ability to avoid detection. It is imperative that annual studies be conducted on Turnitin and other such packages to assess their ability to detect AI plagiarism while minimizing false positives and keeping false negatives at a reasonable level. Perhaps, such assessment studies need to be supported by organizations such as the National Science Foundation (NSF) and conducted by university faculty concerned about misuse of AI by students in higher education settings.

References

Aicinena, S., & Ziyanak, S. (2021). The Native American contest powwow: Cultural tethering theory. Lexington Books.

Bryant, A., & Charmaz, K. (Eds.). (2007). The Sage handbook of grounded theory. Sage.

Elliott, V. (2018). Thinking about the coding process in qualitative data analysis. The Qualitative Report, 23(11), 2850-2861. https://nsuworks.nova.edu/tqr/vol23/iss11/14

Canisius University (2024, August). AI and society initiative. AI and Society Initiative at Canisius University, https://blogs.canisius.edu/generativeai/ai-and-academics-home/ai-and-academics-ai-in-assignments/

Corbin, J., & Strauss, A. L. (2015). Basics of qualitative research techniques and procedures for developing grounded theory (4th ed.). Sage.

Guerro-Dib, G. J., Portales, L., & Heredia-Escorza, Y. (2020). Impact of academic integrity on workplace ethical behavior. International Journal for Educational Integrity, 16(2), 1-18. https://doi.org/10.1007/s40979-020-0051-3

Halbert, D. J., DiMatto-Gibson, D., Cabrera, M., Mazurowski, T., Ingram, M. (2025). Artificial intelligence and plagiarism. Online Journal of Distance Learning Administration (OJDLA), Vol:28 (1) p. 1-15.

Ma, K., & Ballardo, J. (2025, October). Berkeley social sciences explores the promise and pitfalls of AI tools. UCBerkeley, https://ls.berkeley.edu/news/berkeley-social-sciences-explores-promise-and-pitfalls-ai-tools

Oravec, A. J. (2023). Artificial intelligence implications for academic cheating: Expanding the dimensions of responsible human-ai collaboration with chatgpt and bard. Journal of Interactive Learning Research, 34(2), 213-237. https://philpapers.org/versions/ORAAII

Perkins, M., Roe, J., Postma, D., McGaughran, J., & Hickerson, D. (2024). Detection of gpt-4 generated text in higher education. Combining academic judgement and software to identify generative AI tool misuse. Journal of Academic Ethics, 22(2024), 89-113. https://doi.org/10.1007/s10805-023-09492-6

Soyer, M., & Ziyanak, S. (2018). The battle over frac(k)ing: The Mobilization local residents. The Qualitative Report, 23(9), 2222–2237. https://doi.org/10.46743/2160-3715/2018.3248

Strauss, A., & Corbin, M. C. (1998). Basics of qualitative research: Grounded theory procedures and techniques. Sage Publications.

Turn-it in. (2025). https://www.turnitin.com/press/turnitin-named-overall-edtech-company-of-the-year

Walters, H. W. (2023). The effectiveness of software designated to detect Ai-generated writing: A comparison of 16 Ai text detectors. De Gruyter Open Information Science, 7,1-24. https://doi.org/10.1515/opis-2022-0158?urlappend=%3Futm_source%3Dresearchgate

Weber-Wulff D., Anohina-Naumeca, A., Bjelobaba, S., Foltynek, T., Guerrero-Dib J., Popoola, O.,  Sigur, P., & Waddington, L. (2023). Testing of Detection Tools for AI-Generated Text.  International Journal for Educational Integrity, 19(26), 1-39. https://doi.org/10.1007/s40979-023-00146-z