During the first half of 2025, the first Generative Tutoring Systems (GTS) began to be explored and tested in European universities and educational startups as part of pilot initiatives. Unlike traditional adaptive systems, these tutors rely on Large Language Models (LLMs) to generate dynamic explanations, personalized examples, and instant feedback, all in natural language.
What distinguishes them from MOOCs?
Unlike traditional MOOCs, which follow pre-designed learning paths, GTS courses don't use fixed flows. They utilize generative AI with prompts that adapt in real time to the learner's level, previous answers, and learning style.
For example, if a student makes a mistake in algebra, the tutor not only points out the error, but can also generate new examples, analogies, alternative explanations, and even small interactive surveys to reinforce the concept from different angles.
Immediate benefits
→ Personalized attention: tailor-made teaching, adjusted to the student's pace and needs.
→ Constant adaptation: without the need for prior programming, the tutor automatically adjusts the difficulty.
→ Linguistic accessibility: explanations in multiple languages and with different levels of formality, which can help break down cultural and pedagogical barriers.
Technological implementation
Companies like LearnGenAI and TutorGPT are exploring and piloting these systems in universities and corporate environments. The architecture of GTS typically combines:
→ A static knowledge base (syllabi, examples, exercises).
→ Adaptive components that analyze student performance and errors.
→ Real-time prompt generation based on context.
→ Automatic creation of summaries, interactive tests and reinforcement paths.
Challenges and obstacles
→ Academic validation: it is necessary to ensure that the generated answers are aligned with pedagogical criteria and free of errors.
→ Bias in education: it is important to avoid stereotypical or culturally biased content based on region, gender or learning style.
→ Continuous assessment: it is necessary to measure the real impact on the student's understanding, retention and motivation.
And the future?
If these systems are widely adopted, we could see hybrid classrooms where an AI tutor handles between 50% and 70% of individual feedback, freeing up teachers to focus on developing socio-emotional skills, critical thinking, or group dynamics. This projection is based on current trends and the potential of the technology, although its widespread adoption will depend on technical, pedagogical, and ethical advancements.
Are we ready for a classroom where the digital tutor works side-by-side with the human teacher? If technology continues to advance with pedagogical responsibility, GTS could mark a key step towards truly personalized and universal education.
We recommend reading our article about LawZero and “Scientist AI”: designing an honest AI to supervise generative agents.