"Prohibiting AI is absurd": academics warn of problems evaluating their students' work

In addition to being a journalism professor, Nancy has spent these last two years having to become a detective in her own classroom. Or, more accurately, a deception disabler, because in every assessment she gives her first-year students, she must design a method to prevent and simultaneously detect the use of the most feared weapon among teachers today: artificial intelligence. Nancy, who preferred not to give her real name, has been teaching written journalism workshops at a renowned private university in central Santiago for four years. Suddenly, especially starting in 2025, the errors in her students' work began to disappear. Nobody submitted texts with spelling mistakes or writing issues. The structures were perfect. And they all looked too similar. She had to change her way of evaluating. The classic format of assigning a week's worth of work for the following week became obsolete: it was evident that most were using ChatGPT or Claude, but the university lacked tools or protocols to detect it. "Now I do a lot of work in class and walk around the computers all the time, checking what tools they use, I stay very alert," says Nancy. "And I also closely monitor what they submit: if last time they handed in a piece of junk and the next one is too good, or vice versa, something is going on. It's suspicious." According to one study, 79% of university students use AI weekly for their academic work. According to a survey by Cadem and Instructure, conducted among university students and teachers in July 2026, 79% of students use AI weekly for their academic work. With this level of adoption, Nancy knows that "prohibiting it is absurd." But she also knows that, as they use it now, "AI makes them skip processes and learning. That is serious because they stop thinking, which generates very uncritical minds." "I tell them a lot, as if we were in school, that an honest piece of work, with errors, is better than something perfect but copied. There are several who aren't even trying, who just want to hand something in and get a good grade. That's why now, when they hand in bad work, I am not as harsh as before. I even congratulate them. Because if they hand me something with many errors, I know they at least made the effort to write it themselves." Reevaluating Assessment Students have always tried to find the shortest path to a good grade. From copying the answers of the student next to them to plagiarizing an essay, the tricks are as old as education. "This is not a new issue," says Antonia Larraín, academic vice-rector at the U. Alberto Hurtado. "What is happening is that now, with AI, the dimension is much greater." When ChatGPT became widespread in the U.S. in 2023 and then in the rest of the world, the first concern appeared in classrooms. When making presentations or essays, students no longer had to Google to find information or even find another text to copy; now, AI, in exchange for a couple of instructions, does the entire job in less than a minute. A suspicion was automatically installed in professors toward their students. Some, without success, tried to prohibit the use of AI. Others try to incorporate it into their classes and tests. Globally known was the case of Roberto Serrano, professor of mathematical economics at Brown University, one of the most prestigious universities on the East Coast of the U.S. In one of his midterm exams, which could be taken at home and submitted online, university students signed a commitment not to use calculators or the internet. The results were extraordinary: nearly half of the students achieved the highest grade (100), and the average for the course was 96. Antonia Larraín, academic vice-rector of the U. Alberto Hurtado. The university graders alerted him to possible use of ChatGPT. Suspicious, the professor decreed that the final exam would be in person and that, if the distribution of grades followed the same pattern as the midterm, only the final grade would count. What happened? The average dropped to 48 out of 100, and only 59 students showed up for the exam. Of the 27 who arrived, 22 had scored a 100 on the previous exam. "Academic integrity," Serrano denounced, "is in danger." "Academic integrity has been a matter all along," thinks Sergio Celis, researcher at the Center for Advanced Research in Education at the U. de Chile. "There have always been copies and terrible cases. AI increases those risks, but in no case does it mean this model collapses. We must think about how to create better assessments so that those risks are minimized." At UAH, they also have this diagnosis and have named it "AI-resilient education." "Other universities are discussing whether to prohibit it or not allow it in the first years," Larraín recounts. "We are not in that position because prohibition generates a dynamic of norm subversion: people use it anyway, but you have to keep catching them. And it doesn't solve the problem, which is dealing with this cognitive technology that is transforming, perhaps like never before in history, the work of a university." The vice-rector believes that traditional university exercises, such as essays, are today "impossible, or you have to do them on-site." Recently relegated modalities are also returning, such as oral evaluations, debates, or handwritten work. Rolando de la Cruz, director of the AI diploma at the U. Adolfo Ibáñez. "These are methods that gain special relevance in this new scenario," says Rolando de la Cruz, academic director of the AI diploma at the Faculty of Engineering and Sciences at the U. Adolfo Ibáñez. "But the central concern should not be only determining if a student used AI, but establishing if they truly learned, if they understand what they present, and if they are capable of defending their conclusions with their own arguments. That is the real challenge we face." Crisis or Opportunity? The paradox is that just as 8 out of 10 university students, according to the Cadem-Instructure survey, recognize frequent use of AI for their academic obligations, nearly the same number (77%) are concerned that it might give them incorrect information. Furthermore, 72% believe it could cause them to lose critical thinking, and half think it will be a threat to their future careers. Is AI like a drug that, despite knowing its negative effects, is very hard to stop consuming? According to Sergio Celis, who is also a professor at Beaucheff, there is a "critical adoption by students," who, while exploring the possibilities of this technology, "are waiting for the university and the teaching teams to help them on how to use it." With two goals, the researcher says: "so that it does not harm their own learning, and they can develop their competencies, and so as not to violate established norms." 72% of university students believe that AI could cause them to lose critical thinking, and half think it will be a threat to their future careers. "Students are worried about their learning," adds Larraín. "Even more than one would think." The ball is, therefore, in the court of the universities, which must be able to design "resilient pedagogical devices" under a tight deadline, while also training their teachers and regulating the use of AI. UAI, for example, approved its Institutional AI Policy in 2025, which establishes principles of supervision and human responsibility, transparency, ethics, academic integrity, and information protection. "The values at stake are intellectual honesty, responsibility, autonomy, effort, and commitment to knowledge," says De la Cruz. "But also something deeper: our capacity to form people who think for themselves." Sergio Celis, academic and researcher at the CIAE of the U. de Chile. Although it may sound sentimental, for Celis, this crisis is an opportunity to reformulate evaluation processes. "We have seen the emergence, prompted by AI, of evaluations where what is interesting to measure is not a final result, whether a thesis or an exam, but the student's progress. Through drafts or advances, one can see improvements. The emphasis is placed more on the rate of change than on the grade." "I am confident that the eruption of AI will create an urgency for the changes required within the classroom," adds the UAH vice-rector. "How to give more relevance to evaluations but also to put students in a much more active role, and for teachers to assume that not all learning depends on what we talk about. We need a system that forces students to think, to have cognitive friction. That they enjoy the difficulty." What about online education? "It is often said that the online modality faces more risks in evaluation and academic integrity, but it is precisely the opposite," says Adriana Caballero, director of Educational Experience and Innovation at the professional institute IACC, whose programs are 100% online. "In a digital environment, we can track every student interaction and see how they progress. That data allows us to adapt to a rapidly changing scenario and accompany each student in what they seek: strengthening their competencies for work." For them, "a multiple-choice questionnaire was never enough, and AI only made that evident." For this, they are creating new ways to evaluate, "because we do not design for the cheater, but for the evaluation to be a formative experience. The goal is not to turn the professor into an AI detector, but to give them tools to understand its capabilities and limitations and recognize when its use is supporting learning and when it is replacing it."