The potential

AI tutoring tools — which provide personalised explanations, immediate feedback and can adapt to individual learning pace — offer possibilities that have long been pursued in educational technology. Research on one-to-one tutoring shows significant learning benefits; AI tools can deliver some of the feedback and personalisation advantages of individual tutoring at scale. Tools including Khanmigo (Khan Academy's AI tutor) have shown promising early results for supporting mathematics learning.

The academic integrity challenge

The most immediately disruptive impact of generative AI in education has been on academic integrity. AI-generated text is now sufficiently fluent that it is extremely difficult to detect reliably; AI detection tools produce false positives (flagging human writing as AI-generated) and false negatives (missing AI-generated text) at rates that make them unreliable for disciplinary purposes. Universities and schools are in the process of restructuring assessment: oral exams, in-person closed-book assessments and portfolio-based assessment that captures process as well as product are receiving renewed interest.

How it affects learning

The effect of AI assistance on learning outcomes depends critically on how it is used. Students who use AI to complete tasks for them — using ChatGPT to write their essay, then copying the output — do not develop the underlying understanding. Students who use AI as a thinking partner — generating initial ideas and then critiquing them, using AI explanations to identify their own gaps, asking AI to generate practice problems — can develop deeper understanding. The design of learning activities matters enormously.

The equity dimension

AI educational tools require reliable internet access, capable devices and the digital literacy to use them effectively. In contexts where these are not available — many lower-income schools globally, and pockets of disadvantage even in wealthy countries — AI tools exacerbate rather than reduce educational inequality. Training teachers to use and critically evaluate AI tools is also unevenly resourced.