Generative AI has many of us scurrying to ensure that our students’ work is their own, especially as the large language models grow more sophisticated. Pinpointing academic dishonesty has always been tricky, but as we begin a new academic year, we will likely face the challenge of verifying authenticity in our business communication students’ work frequently.

Some who have experienced this situation suggest we start with a conversation that first explains why it is important to be able to write without the aid of AI as well as reminds students about academic integrity and your institution’ s plagiarism policies. After that, you can move into showing the student why you suspect inappropriate use of AI.

Becoming aware of AI’s unique tells can help with this onerous chore. The tips below are indicators of possible AI plagiarism.

1. Assignment is overly long. If you have asked for a short e-mail and the student turns in a lengthy treatise, it’s time to ask questions.

2. Work omits key elements of the assignment and submission is too long. If you’ve asked your business students to write a short report and it’s not only longer than the assignment requires but also does not include specifics you’ve asked for (i.e. relevant footnotes/endnotes or other important components), you have reason to be wary.

3. Author uses trite expressions. Beware of cliché use. AI writing relies heavily on these overused phrases that are ubiquitous on the Internet.

4. Student surprisingly submits the assignment early. Some students are eager to turn in their work before it is due, but be leery of those who have never shown this inclination with prior assignments.

5. Prose relies too much on lists and bulleted points. Generative AI regularly displays responses using lists and bullets, but most humans do not. Unless you have specifically asked for students to organize their work this way, using bullets and enumeration can indicate AI plagiarism.

6. Assignment contains no writing errors. Few college students are capable of producing error-free prose.

7. Composition uses overly complex sentence construction. If AI has been used, this is not a fluke, since AI generates text from a mixture of styles or genres.

8. Assignment doesn’t match a student’s previous work. Writing instructors are uniquely qualified to identify students’ work. Any assignment that shows massive improvement over previous assignments is a red flag that AI may have been involved.

Additionally, AI prose has its own characteristics including repeating ideas and phrases; using the same sentence structure throughout; including inaccurate facts and claims (hallucinations); and reliance on generic explanations without supporting details. AI also often uses outdated content or shows a lack of expertise when discussing complex ideas, especially those requiring first-hand knowledge.

Likewise, be on the lookout for AI’s favorite words (an incomplete list!)

• Consequently
• Crucial
• Delve
• Dive
• Furthermore
• In today’s [adjective] world
• Not only but
• Tapestry

Conversely, students commonly choose words that AI does not. These include anything, because, believe, feel, nothing, if, since, so, thing, think, and too. Finally, AI prose lacks voice, that illusive element of writing that makes it sound like it came from a human.

Detecting AI use in student writing is a burdensome task, and one that shows no signs of becoming easier as the tools become more sophisticated.

Bowers-Abbot, M. and Abbot, W. (2024, Feb. 7.) Human or AI? Connectives hold the clues. Faculty Focus. https://www.facultyfocus.com

n.a. (2023, June 28.) AI text detection. ChatGPT