Students who use artificial intelligence (AI) to complete assignments may perform well while the technology is available but struggle to apply what they have learned independently, according to a new review by researchers at Stanford University. The findings suggest that AI-assisted learning produces better results when students use the technology to understand concepts and practise skills rather than generate answers or complete academic work for them.
The November 2026 edition of The Evidence Base on AI in K-12, published by Stanford University's SCALE Initiative on November 10, examines 52 causal research papers on the use of AI by students and educators. The evidence base has more than doubled since the previous review in March 2026, which examined 20 studies.
The latest update includes 32 additional studies published between October 2025 and June 2026. Of these, 28 examined AI tools used directly by students, while four focused on tools designed for educators. Five of the newly included student studies were conducted in US K-12 settings. However, researchers cautioned that the findings cannot yet be generalised across all schools, as much of the available research continues to focus on higher education.
The review found that students who use AI to outsource academic tasks can face difficulties when they later have to complete similar work without technological assistance. In one study involving Brazilian undergraduates, AI helped students prepare a presentation more quickly, but their performance declined in a subsequent conceptual assessment. Another study involving Canadian undergraduate students found that beginners who used AI to generate complete coding solutions could finish programming tasks but demonstrated limited understanding of the solutions and the concepts behind them.
An analysis of student activity on an adaptive mathematics platform in the US also found that, following the release of ChatGPT, students spent less time on text-based practice problems that could easily be entered into a chatbot than on graph-based questions. Their performance on proctored assessments also declined on the text-based problems relative to graph-based ones.
The researchers noted that students may have used AI to bypass some of the problem-solving involved in practice exercises, leaving them less prepared to tackle similar questions independently.
The review, however, found more favourable, though inconsistent, results when students used AI to learn unfamiliar material, receive explanations, practise problems and obtain feedback on their own work. Some studies reported improvements in learning, while others found no significant advantage over conventional teaching and practice methods. The researchers also found that positive student experiences with AI did not always translate into better academic performance.
The findings follow a separate study of cognitive offloading by researchers at Stanford's Graduate School of Education. Cognitive offloading refers to using external tools to reduce the mental effort required to complete a task. While this can help people manage routine or repetitive work, researchers caution that delegating the thinking involved in learning can reduce opportunities to develop independent skills.
In an experiment involving Brazilian middle-school students, researchers divided participants into four groups with different levels of access to an AI tool and asked them to complete two creative tasks. Students who used AI during the first task lost their performance advantage when the tool was withdrawn. Those who had relied more heavily on AI performed worse on the subsequent creative task than students who had never received AI assistance.
“If I can completely delegate a task to AI and I never spend time thinking about it, then there’s no learning,” said Guilherme Lichand, Assistant Professor at Stanford Graduate School of Education.
Lichand said the ability to persist through difficult tasks and regulate one's learning process was an important part of developing knowledge and applying it to unfamiliar situations. The researchers also raised concerns that excessive reliance on AI could affect students' confidence in their own creative abilities and their capacity to work through academic challenges.
Lily Fesler, Senior Research Associate at Stanford's AI Hub for Education, said the impact of AI depended on the type of tool and how students used it. “What we’ve learned so far is that it has the potential to both improve and harm learning outcomes. The difference likely comes down to the type of tool that’s being used,” she said.
The review also examined how AI affects teachers and classroom instruction. Earlier studies found that AI could reduce the time teachers spent preparing lessons without lowering the quality of their work. AI-generated suggestions and drafts also helped educators provide more detailed feedback to students without increasing the time spent on the task.
However, one newly included study found that providing teachers with an AI-based support tool reduced students' motivation and confidence. The review also found that the effects of AI assistance could vary according to teachers' experience and prior performance.
Researchers said AI tools that guide students through a problem, ask questions instead of immediately providing answers, or offer feedback on written work without generating the text themselves could help retain student involvement in the learning process.
Fesler said AI assistants could provide more frequent and immediate feedback on essays than teachers might have time to offer, provided the tools did not write students' assignments or entirely replace teacher feedback.
The review also cautioned that evidence on the long-term effects of AI in education remains limited. While some studies found improvements in short-term learning and retention, researchers said there was insufficient evidence to establish whether these gains persist over longer periods.
Stanford researchers have called for more controlled studies and a measured approach to introducing AI into classrooms, particularly as students and teachers increasingly use the technology for academic work.