'I can't do assignments without AI': Why banning chatbots misses the real problem

Photo not related to article content. [Getty Images Bank]

"I can't even start an assignment without AI." "My hands shake when the chatbot goes down." "I only feel at ease after I've asked AI first."

These are not unusual things to hear on college campuses. Some students open an AI tool before anything else when faced with an assignment. Others, given the same task, barely reach for it at all. What drives that difference?

The easy answer might be that AI is simply convenient. But research points elsewhere. The reason some students become dependent on AI has less to do with convenience than with how much they believe in themselves.

A study published in the July 2026 issue of Frontiers in Psychology by Cai Zhang, a professor at Jinling Institute of Technology in eastern China, found that dependence on generative AI grows stronger the less confident students feel about their own academic abilities.

'AI is convenient' — only half the story

Generative AI has spread rapidly across university life. Students routinely turn to chatbots to search for information, draft writing and work through problems.

The concern is that a growing number of students are moving beyond treating AI as a tool and into a state of anxiety when it is unavailable. That is precisely what the research team set out to examine — why some students come to lean on AI while others do not.

AI apps, including Claude, displayed on a screen. [Getty Images Bank]

The research team surveyed 428 students at five universities in eastern China in two rounds, three weeks apart. The first round measured confidence and stress levels; the second assessed students' perceptions of AI and how dependent on it they had become.

Four variables were measured: belief in one's ability to succeed academically, stress arising from studies, perceived usefulness of AI and degree of AI dependence. Each item was rated on a five-point scale.

The first variable is known in academic terms as "academic self-efficacy" — in plain language, confidence in one's own ability as a student: the degree to which a person believes they can get through exams and difficult assignments on their own.

Students with lower self-belief showed higher AI dependence. Those who doubted their ability to complete assignments on their own turned to AI for instant answers. Students with higher confidence, by contrast, tended to work through problems themselves and used AI less.

Notably, factors such as gender, year of study, type of institution and academic major showed little relationship with AI dependence. It made no difference whether a student was male or female, a first-year or a senior, enrolled at a selective or an ordinary university, or studying science or the humanities.

The assumption that science and engineering students use AI more did not hold up, nor did the idea that more advanced students would be more inclined to rely on it.

The tendency to depend on AI came down to how students assessed themselves. If two students sitting in the same classroom showed different levels of AI dependence, the gap was explained not by their major or year of study but by their confidence.

'Build confidence,' not 'ban AI'

The research team said blanket bans on AI use are not the answer. Providing appropriate academic support and step-by-step feedback — helping students feel they can manage on their own — reduces the tendency to lean on AI, they said.

Reframing how students think about AI was also proposed: teaching them to treat it as a tool that assists independent thinking rather than replaces it. The team suggested incorporating into curricula lessons on critically engaging with AI and developing the habit of reviewing AI-generated results.

The team acknowledged that the study was limited to Chinese university students, and said follow-up research is needed because results may differ in countries with different entrance-exam pressures and educational environments. They added that the study did not measure in detail how frequently or in what ways students actually used AI.

Reference

DOI: 10.3389/fpsyg.2026.1784354

Zhang C (2026) The relationship between academic self-efficacy and GenAI dependence among college students: a chain mediation model of academic stress and perceived usefulness of GenAI. Front. Psychol. 17:1784354.

dbsdn1110@heraldcorp.com