Do Artificial Intelligence Tools Really Improve University Students’ Results?

The use of artificial intelligence tools capable of generating text, images, or even code is gradually transforming higher education. A recent study analyzed how students from six different fields, across two universities, integrate these technologies into their academic work. The researchers compared the performance of groups allowed to use a conversational AI assistant with those limited to traditional resources such as textbooks or lecture notes.

The results show that students with access to the tool generally achieve better scores, especially in subjects like computer science, systems administration, or childhood disorders. This improvement is particularly noticeable when students interact actively and thoughtfully with the tool, for example by rephrasing their requests to refine the responses. A positive, albeit modest, correlation was observed between the number of modifications made to queries and the quality of academic results.

The students surveyed highlight several advantages: quick access to information, help in clarifying complex concepts, and time savings in organizing their work. Some also mention the possibility of receiving alternative explanations when traditional sources are insufficient. However, concerns remain. Errors or inaccurate responses generated by the tool, as well as the risk of excessive dependence, are frequently mentioned. Some also fear that passive use of these technologies may reduce their personal critical thinking efforts.

Students’ expectations also reveal a desire for more interactive and personalized tools. Many would like multimodal features, such as image generation, presentations, or audio explanations, as well as better support for scientific or technical subjects. These requests reflect a growing need to adapt to different learning styles and reduce cognitive barriers.

The study suggests that artificial intelligence can indeed support learning, provided its use is supervised and accompanied by training in critical thinking. The differences in results across disciplines also indicate that its effectiveness depends on the context and how students use it. Finally, these observations open avenues for making education more accessible, especially for students with specific needs, by leveraging the potential of these tools to provide varied explanations and tailored support.

References

Origin of the Study

DOI: https://doi.org/10.1007/s10209-026-01315-w

Title: Expanding the lens: multi-institutional evidence on student use of ChatGPT in higher education

Journal: Universal Access in the Information Society

Publisher: Springer Science and Business Media LLC

Authors: Miguel Á. Conde; Rocío García-Pascual; Francisco J. Rodríguez-Sedano; Jesús-Ángel Román-Gallego