Evidence guide · Education
AI in education can mean very different things: a learner discussing a concept with a chatbot, a teacher preparing draft feedback, or a school using an automated system. A useful assessment starts with the learning task and the people affected. This guide connects the questions behind those uses to our reporting on learning outcomes, teacher competencies, access and AI literacy.
By The Impact of AI Editorial Desk · Published 14 October 2026. Original research dates remain visible in the linked reports.
Possible uses include preparing an initial lesson outline, explaining a concept in another way and giving learners material to question or revise. The important comparison is with the usual teaching method, including the time needed to check and correct an answer. A fluent explanation, a positive survey response or a faster draft does not by itself show that someone learned more.
For each proposed use, identify who checks the output and how a learner can get help from a person. Consider whether the task develops understanding or merely produces the work used to assess it. The same tool can support practice in one setting and conceal a gap in understanding in another.
Look for studies that actually test learning: what learners understood before and after an intervention, how long the effect lasted and what comparison group received. Check whether the evidence concerns a specific AI tutor, a chatbot conversation or digital teaching more broadly. These are different interventions.
Our reporting includes a meta-analysis whose digital-teaching evidence should not be read as an AI-specific effect. A headline average can also hide differences in age, subject, language, access and teacher support. Read the study’s scope and limitations before applying a result to a classroom.
AI literacy includes recognising uncertainty, checking sources and understanding when a tool is appropriate. It is broader than learning a particular prompt. Teacher competency frameworks and assessment discussions can help organise those responsibilities, but a framework is not proof that a product improves learning.
Ask what work learners must do themselves and how they should disclose assistance. A useful assessment should reveal what a learner understands. Include examples where an AI answer is plausible but wrong, and evaluate the reasoning used to check it rather than only the polish of the final response.
Availability does not imply equal support. A learner with a paid account, a reliable device and someone to check answers may have a different experience from a learner without those resources. Evaluate the support around the tool alongside its capabilities.
Before a classroom trial, establish the data needed, the alternatives available and how errors or concerns will be handled. Keep identifying or confidential material out of informal tool experiments. The reporting below provides evidence and context; a local school’s responsibilities and decisions require their own review.
Potential benefits depend on the task and the support available. Risks include inaccurate explanations, over-reliance, unequal access and privacy concerns. Judge a proposed use against measured learning, teacher oversight and a clear alternative.
An AI task result does not establish that a system can take responsibility for teaching, safeguarding, assessment and a learner’s wider needs. Examine the specific task and the evidence rather than extrapolating from one demonstration.
Read our source analysis and its limitations, then inspect the original evidence. A framework, a survey, a technical benchmark and a clinical outcome study answer different questions.
A UK survey records near-universal student use while a Japanese public training listing shows AI entering vocational education. The shared challenge is turning access into genuine learning rather than unexamined dependence.
A new UNESCO IITE analysis compares national, institutional and academic frameworks for teacher AI competence. It offers a policy checklist, but does not test whether any framework improves classroom outcomes.
Read the evidence analysis · Original source: UNESCO Institute for Information Technologies in Education
Teachers need to know what helps a learner, not just what looks promising in a headline. A review of 477 learners favours technology-supported teaching, but only a few of its studies explicitly tested AI.
Read the evidence analysis · Original source: Frontiers in Education
The OECD's learner framework describes the knowledge, practices and values students need to understand AI systems, work with them critically and recognise their social and environmental consequences.
UNESCO
AI-generated illustration · not a video frame
YouTube loads only when you choose. Press play in the player to start. If playback is unavailable, use “Watch on YouTube” below.
Education & human agency · Background explainer
Published 2025-01-23 · 8 min 30 sec
A conversation about how teachers can guide AI use while preserving pupils’ judgement. Watch for the difference between helping a learner think and supplying an answer.
Evidence check: Teachers’ perspectives provide context. They do not establish that a particular AI tool improves learning outcomes.
Source checked 14 October 2026. Check available captions on the publisher’s video page.