AI Is Changing Higher Education — Just Not the Way We Expected

What a 2026 study on AI-integrated learning teaches us about guidance, not shortcuts

Not as Obvious as Expected

Artificial intelligence (AI) is changing higher education. I’ll admit, it’s not been as obvious as I thought it would be, but the changes are there. In truth, on the surface, not much has changed. Some institutions are on board and moving forward fast, others started strong, and others are still not quite moving at full speed. This is also true for faculty.

In the meantime, our students have been using it in higher numbers. Those without guidance are using it in ways that are less than ideal for their learning and the value of their program. On the other hand, those who are being encouraged to use it and are guided as they do so, as I expected years ago, are doing very well. For a while now, the most important question has not been whether AI belongs in the classroom, but how it can be used to improve student learning.

What the Study Found

A 2026 study by Yermaganbetova et al. is a great example and evaluation of an AI-integrated learning platform in a 3D modeling course. The researchers found that students using the AI-supported platform performed better on post-tests, completed practical tasks more successfully, and reported higher levels of motivation, engagement, autonomy, and self-regulated learning than students in a traditional instructional setting.

The important thing here is that the findings show that AI is beneficial when it acts as a guide rather than a shortcut, which is not a new story, I admit. The platform in the study didn’t replace the instructor but gave students recommendations, timely feedback, and automated comments that helped them work through difficult concepts. Basically, the AI supported learning by helping students monitor their progress, revise their work, and stay engaged long enough to improve. This is important for anyone studying or teaching in this area. The problem isn’t that AI exists, but whether faculty and institutions let it replace the thinking they are supposed to cultivate.

From a student learning perspective, the article suggests that AI can strengthen both outcomes and habits. Students not only scored better, but they also showed more self-regulation. This is an important finding because learning isn’t just about getting the right answer once on a test (and if it is, we should probably ask whether that test is worth giving); it’s about developing the ability to plan, monitor, and adjust thinking while, hopefully, understanding the practical foundation of the learning. When AI helps students do those things, it becomes part of the learning process rather than a shortcut to get around it. This is such an important point for faculty who worry that AI encourages dependency. The evidence from Yermaganbetova et al. (2026) suggests the opposite can happen if the tool is designed and used appropriately.

Design Matters More Than the Tool

However, even with the positives, the study also highlights a bigger challenge: AI is not inherently beneficial. The results depend on how courses are designed and how faculty present the use of AI, which is why practical solutions become essential. One solution is to build AI into assignments that require reflection, not just production. For example, students can be asked to explain how they used AI, what the tool contributed, what it missed, and how they revised the output. This approach shifts the focus away from speed and artifact production toward demonstrating understanding of the process and its importance. It also helps instructors evaluate whether students are learning or just outsourcing the work.

A second solution is to train faculty to use AI as a pedagogical partner rather than a threat. Many instructors still view AI as something they must either prohibit outright or tolerate as nothing more than an invasive species, feeling they have no choice. A better approach is to treat it as an instructional design issue. Faculty development can show instructors how to use AI for formative feedback, adaptive practice, example generation, and progress monitoring while still preserving academic integrity. The study’s findings suggest that when AI provides timely, specific feedback, students are more likely to stay engaged and build autonomy. That means the faculty role changes, but does not disappear. Teachers become designers of learning conditions, not just deliverers of content.

A third solution is to create clear course policies that define acceptable AI use. Students do better when expectations are clear. In fact, I now have a clear message at the beginning of courses outlining the consequences of academic integrity concerns and have seen a sharp drop in issues. It’s important that students know whether AI is allowed for brainstorming but not for final drafting, or whether it may be used for feedback but not for answer generation; the guidance should be stated early and reinforced often. Ambiguity invites misuse, but clarity leaves no doubts over expectations or consequences.

A fourth solution is to design assessment methods that measure the process. Traditional assignments often reward the finished response and ignore the steps that produced it. Used unethically, AI can take this a step further and remove the process entirely, leaving a polished product that the student has probably not even read. It has their name on it, but they don’t know what it is. If educators want to know whether students can think, we must ask students to show their reasoning through drafts, oral explanations, revisions, and applied problem-solving. This type of assessment is harder to fake and more aligned with the kind of learning the Frontiers study supports: active, self-regulated, and skill-based learning.

Rethinking the Product Itself

Another very important tangent comes from this: the product. If we simply ask students to write a paper, what is the purpose of writing the paper? Has the student experienced anything? Has anything improved in the world because of it? We have a real opportunity here to take things to the next level. With the tools available, we can have undergrad students learning through application. We can have grad students designing solutions to problems. We can have doctoral students implementing and measuring the solutions. By encouraging students to learn a faster, but carefully guided, process, we can expect more, they can learn more, and we can make the world a better place at every stage of the academic journey.

AI should not be treated as either an enemy of learning or as a magic solution. It is better viewed as a tool that magnifies whatever design principles guide it. If a course is weak, AI may make the weakness easier to hide. If a course is intentional, structured, and student-centered, AI can help students move deeper into the learning process. That is why institutions should invest not only in AI tools but also in faculty development, course design, academic integrity policies, and student digital literacy. These supports help ensure that AI strengthens education instead of hollowing it out.

In the end, the most promising use of AI in higher education is one that improves the classroom by making learning more responsive, more personalized, and more reflective. The Yermaganbetova et al. (2026) study shows that when AI is used thoughtfully, it can improve both performance and student engagement. The challenge for educators is to create courses where AI supports human learning rather than substituting for it.