How to Quickly Identify AI-Written Essays
A Conversation with an Academic Journal Editor
Last week, I invited the editor-in-chief of a respected U.S. academic journal—Dr. Mark Ellison—to give a guest lecture at our university in Boston. Since I happened to be free that day, I drove with a colleague to pick him up from South Station. During the ride back to campus, we began discussing a growing concern among educators: how to identify student papers written with artificial intelligence.
Based on my experience reviewing undergraduate proposals this semester, I shared several telltale signs.
1. Unusually Long Essays
AI-generated essays are often excessively long. As Dr. Ellison joked, “Even if you ask ChatGPT a nonsense question, it’ll give you a full-blown thesis.” When students don’t understand the topic themselves, they tend to paste large sections of AI-generated responses into their papers. This semester, I received multiple project proposals that were 6,000–7,000 words long—far exceeding the 1,000-word norm in previous years. Whenever I see such unusually lengthy submissions, it raises a red flag.
2. Rigid Structure and Formulaic Language
AI essays typically follow strict patterns such as list-based structures, “two-sides” analyses, or the classic introduction-body-conclusion model. The body sections often present basic, universally accepted knowledge from the field, segmented neatly into dimensions with repetitive tone and sentence structure. Every paragraph reads similarly—flat and expository—and the conclusion usually begins with, “In summary...”
3. Lack of Personal Voice or Naivety
Human-written essays—especially from undergraduates—often include quirky opinions, youthful biases, or even factual misunderstandings. That’s natural and expected. Some students use bold metaphors or unexpected word choices that make you smile. AI, on the other hand, produces emotionally neutral, highly structured content that feels like it was written by a 75-year-old academic summarizing textbook consensus. No passion, no clear opinions, no surprises—just conservative, balanced statements.
4. No Grammar or Punctuation Errors
Another clue is perfection. When I suspect a student used AI, I copy their paragraph into an AI tool to check for typos, poor phrasing, or punctuation issues. If the entire essay—thousands of words—is grammatically flawless, that’s a major sign. Most students, no matter how strong, will make some stylistic or mechanical errors. AI doesn’t.
5. Space Before and After Numbers
One subtle but consistent AI habit is inserting spaces around numeric symbols. For example, “in 2024 ,” instead of “in 2024.” When I find extra spaces around multiple figures in a student’s work, it’s often an indicator that the content was generated by AI.
6. Flat Emotional Curve
Human writing tends to show emotional and rhetorical variation. We emphasize, we question, we surprise. If you map out emotional intensity over time, human writing creates an irregular pattern. AI-generated content, by contrast, is smooth and mechanical. There are no peaks of excitement, confusion, or conviction—it’s all evenly paced and restrained.
Final Thoughts
With the rise of generative AI tools like ChatGPT, Bard, and Claude, educators must become more adept at spotting synthetic writing. As I told Dr. Ellison during our ride: detecting AI isn’t just about catching cheaters—it’s about preserving the value of authentic, personal thought in academic work.
By paying attention to structural patterns, tonal consistency, linguistic perfection, and the absence of human quirks, we can often determine whether a piece of writing was generated by AI. And when in doubt, I often run it through tools like PPTDetector.com, which specializes in identifying AI-generated PowerPoint slides—offering a model for future detectors focused on text analysis.