Generative AI has rapidly become part of everyday student life. Learners use it to understand unfamiliar concepts, brainstorm research questions, improve study plans, receive feedback, and prepare for assessments. Used responsibly, AI can give students something that is often difficult to obtain in a large class or online course: immediate, personalized support. A learner who does not understand a statistical concept, for example, can ask for a simpler explanation, a practical example, and a short knowledge check.
However, the same technology can produce complete essays, solve problems, and generate convincing but potentially inaccurate references. This creates an important distinction. AI can either support the learning process or replace it. The most useful question for educators is therefore not whether students should use AI tutoring. It is how learning activities should be designed so that AI strengthens independent thinking.
Emerging research illustrates why Instructional Design matters. A 2025 study published in Scientific Reports examined a carefully designed AI tutor in an undergraduate physics course. Under the conditions studied, students learned more in less time and reported stronger engagement than students attending an in-person active-learning class. This does not mean that any chatbot automatically improves learning. The AI tutor was structured around defined learning objectives, appropriate scaffolding, and established teaching principles.
Another large field experiment on generative AI and learning found a different result. Students with access to an unrestricted AI assistant performed better during practice, but some performed worse when the AI was removed. A more carefully designed AI tutor with learning safeguards reduced this negative effect. Together, these findings suggest a practical principle: AI tutoring is most valuable when it behaves like a tutor rather than an answer machine.
The LEARNT model is a six-step framework that educators, online tutors, and learning designers can use when incorporating AI tutoring into academic and professional learning.
Before opening an AI tool, students should identify what they are expected to learn. An assignment might require a student to evaluate competing theories rather than simply describe them. A dissertation task might require the learner to develop a defensible research question. A professional safety activity might test whether someone can identify hazards and justify suitable controls. Without a clear learning outcome, students can easily focus on producing polished text instead of developing the required knowledge. Educators can support this step by translating assessment requirements into clear questions:
Students should make an initial attempt before asking AI for help. This might be a short outline, an explanation in their own words, a proposed research question, or an initial solution to a problem. The first attempt does not need to be perfect. Its purpose is to activate prior knowledge and expose gaps in understanding. A student planning a dissertation, for example, could first write:
The student can then compare this thinking with the feedback provided by AI tutoring. Starting with AI-generated content makes it harder to determine what the learner genuinely understands.
The quality of an AI-supported learning activity depends heavily on the type of assistance requested. Prompts that ask AI to produce a complete assignment remove the intellectual work that assessments are intended to measure. Coaching prompts are more valuable because they keep the student responsible for the final decisions. Useful prompts include:
These instructions position AI as a questioning partner, formative reviewer, or practice tutor.
AI-generated information should never be treated automatically as reliable. Language models can produce incorrect facts, outdated guidance, distorted summaries, and references that do not exist. Students should verify important claims using original, credible sources. For dissertation and thesis research, this normally means consulting academic databases, peer-reviewed publications, official statistics, professional standards, and primary documents. A responsible verification process should include the following questions:
This step also develops information literacy, which is increasingly important as AI-generated material becomes harder to distinguish from verified knowledge. The UNESCO AI Competency Framework for Students similarly emphasizes critical judgment, ethical awareness, and responsible participation in AI-supported environments.
Students should record how AI contributed to their work. A simple decision log can include:
Documenting the process discourages uncritical copying and makes student reasoning more visible. It can also help supervisors provide better feedback because they can see how the project developed. Institutions should give students clear instructions about when AI disclosure is required. Expectations should be specific to each activity because acceptable support during brainstorming may not be acceptable in a final examination or assessed report.
The final test of AI-supported learning is whether the student can perform without AI. After using AI to study a topic, learners could:
If a student can produce an impressive document with AI but cannot explain its reasoning, meaningful learning has probably not occurred.
Consider a postgraduate learner developing a dissertation research question. The learner begins by defining the intended learning outcome: producing a focused and researchable question supported by a clear academic rationale. The student then writes an initial question independently. Instead of asking AI to select a topic or write a proposal, the student asks it to identify ambiguity, hidden assumptions, variables that require definition, and practical limitations.
The student searches appropriate academic databases and checks each suggested issue against published literature. After revising the question, the student records which changes were made and why. Finally, the learner explains the research question, methodology, and expected contribution without using AI.
The intellectual ownership remains with the student, while AI supplies structured formative feedback. The same process can support individual assignments, literature reviews, thesis planning, workplace learning, and professional qualification preparation.
Telling students to "use AI responsibly" is too vague. Effective online courses should provide practical boundaries and learning structures. Course designers can help by:
Educators should also consider accessibility and digital inequality. Some students have access to advanced paid systems, while others rely on limited free tools. Essential learning outcomes should not depend on purchasing a particular AI product.
AI tutoring can help students improve academic performance, but better grades should result from better understanding rather than automated completion. The most effective AI-supported learning asks students to think first, question actively, verify evidence, document decisions, and demonstrate that they can perform independently. When these safeguards are built into eLearning, AI becomes more than a shortcut. It becomes a tool for feedback, reflection, practice, and intellectual development.