Schools are under pressure to integrate AI, with teachers, students and parents struggling to keep up. The question for many persists: Where do we start? The GenAI4ED project set out to find evidence-based answers.
Since January 2025, Trilateral Research, GenAI4ED’s project coordinator, has led 4 research tasks, about Generative Artificial Intelligence (GenAI) in secondary education.
For the project, these research tasks are necessary to help us build the GenAI4ED platform, and to develop the training materials for the project’s Training Hub and Pilot Studies.
But research findings are not only important for the project, they also inform us of the status of GenAI in educations, and they can help increase literacy (what we know about how GenAI works) by educating teachers, parents and students about GenAI uses in secondary education, limitations, and risks. Additionally, these findings can help us make recommendations for nationwide and institutional policies that foster responsible use in all stakeholders.
Want to learn what we found out?
Following is a summary of these tasks and an overview of some key findings deriving from this work:
The tasks:
- A systematic review of GenAI in education– a careful revision of academic literature regarding GenAI in secondary education contexts.
- A taxonomy of GenAI tools that can be used for education– an evaluation of tools.
- A systematic review of policies and guidelines on GenAI in education– a revision of currently available guidance.
- Interviews on the psychological factors in GenAI-human interaction– learning how stakeholders feel and experience GenAI.
Systematic review of GenAI in education
GenAI4ED members: Alba Paz-López (Trilateral Research-Task leader), Emmanouil Kritikos (Cyprus Institute), Francesca Aloatti (Wonderful Education), and Martina Manzotti (Wonderful Education).
We reviewed 59 academic papers following a PRISMA protocol (a systematic procedure of evaluating publications), to explore the topics:
- Teacher and student perceptions on GenAI in education.
- Factors associated with the effective use of AI in education.
- How educators and other stakeholders choose and evaluate GenAI tools.
- Tasks where GenAI can add value.
- Impact of GenAI on educators’ wellbeing and working conditions.
Key findings:
- GenAI is perceived as beneficial for learning and teaching, assisting in tasks like feedback, personalisation and content creation. It is described as engaging, fun and accessible or easy to use. However, concerns over plagiarism are still very prominent for both teachers and students, along with data privacy, cognitive impairment and dependency concerns.
- Prompt engineering practices (the instructions we give GenAI) and AI literacy are perceived as the most impactful factors in effective GenAI deployment.
- Stakeholders evaluate and choose tools by reviewing their performance in completing tasks, comparing different tools or models, or testing learning outcomes after a period of experimentation.
- Assessment and content creation are the top tasks for AI-assisted teaching. Writing assistance is the most benefited learning task.
Taxonomy of GenAI tools that can be used for education
GenAI4ED members: Pinelopi Troullinou (Trilateral Research-Task leader), Valeria Filippou (Cyprus Institute), Simone Favale (Wonderful Education), Francesca Aloatti, Martina Manzotti (Wonderful Education).
We carefully reviewed 28 GenAI tools that show promise for education, considering the following aspects:
- Educational affordances (how they can help with education).
- User role (target stakeholders).
- Domain of use (humanities, science, etc.).
- AI applications (LLM, NLP, TTS, etc.).
- Type of integration (whether it is a platform, an app, etc.).
- Key risks
Key findings:
- Six GenAI tools emerged as the most valued in the taxonomy, chosen to be featured in the GenAI4ED platform and piloting studies: Canva, ChatGPT, Eduade.ai, Magic School, Microsoft Copilot, Midjourney.
- No tool is free of risks, particularly inaccuracies and intellectual property risks.
Systematic review of policies and guidelines on GenAI in education
GenAI4ED members: Alba Paz-López (Trilateral Research-Task leader), Aine Sperrin (Trilateral Research), Myrianthi Hadjicharalambous (Cyprus Institute), Stefania Aceto (Wonderful Education), Francesca Aloatti (Wonderful Education).
We reviewed 72 guiding documents about GenAI from across the world, focused on secondary education, seeking:
- Prescriptive texts directed at teachers, students, or parents and guardians.
- Policies, guidelines and strategies
Key findings:
- Most documents consist of guidelines, not hard policies or strategies.
- Guidelines highlight concerns, with a particular focus on GDPR (data protection), but illustrate no concrete steps on safe practices.
- Guidelines for students focus on the “dangers of plagiarism”; guidance for teachers emphasise the need for literacy; guidelines for parents underscore the importance children’s safety.
- Scarcity of guidance on specifically how to leverage GenAI for learning or teaching.
- No mention of the EU AI Act[1] (the first European regulation on Artificial Intelligence) in any of the documents.
Interviews on the psychological factors in GenAI-human interaction
GenAI4ED members: Alba Paz-López (Trilateral Research-Task leader), Mallari Shroff (Trilateral Research), Anastasia Kordoni (Trilateral Research), Arja Krauchenberg (European Parents’ Association), Persia Xenopoulou (P.G.M.S.), Tonia Galati (P.G.M.S.), Alexis Mikellides (P.G.M.S.), Dimitra Dimitrakopoulou (Ellinogermaniki Agogi).
We interviewed 44 participants, teachers (18), students (15) and parents (11) focusing on the following aspects:
- Learning processes with GenAI.
- Expectations, perceptions and experiences with GenAI for learning or teaching.
- Skills and role developments as a result of GenAI.
- Emotions, trust and confidence in GenAI use.
- Ethical aspects.
Key findings:
- ChatGPT leads as the most popular tool across the board, with social influence being the main criteria in choosing the tool.
- Students learn about AI mainly from peers, teachers from self-exploration, parents learn from work.
- GenAI is perceived to enhance tasks and make them easier and more accessible.
- Inaccuracies are the leading perceived downfall of GenAI tools, and the leading cause of mistrust for users.
- Teachers and parents express high levels of technostress.
- There is a possible association between learning about GenAI from social media, and diminished literacy.
- Students are the most accepting group of GenAI but show the lowest critical thinking and fact-checking practices.
- In the absence of training, teachers form knowledge–sharing communities.
- Parents express a lack of communication from institutions regarding GenAI.
What this tells us
GenAI’s adoption across educational contexts is an uneven process, but research points at literacy (learning how to use GenAI effectively) and critical thinking as the building blocks of responsible adoption. GenAI offers numerous benefits for learning and teaching, and accessible, targeted training emerges as the key to ensuring that we maximise these benefits in education and minimise risks.
We are still learning…
As GenAI evolves, evidence tries to catch up, but there are still many areas that require immediate action:
- Ongoing research work needs to address the evolving impact of these tools along with the needs of students, teachers and parents.
- Continued and comprehensive evaluation of GenAI tools is paramount to ensure that these adhere to the teaching and learning requirements of all users, and to current legislative texts.
- Research needs to inform policies to achieve effective and practical guidance for stakeholders.
At GenAI4ED we continue working towards achieving these goals, to support safe responsible GenAI-assisted learning and teaching.
Author: Alba Paz-López
[1] European Parliament and Council, Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence OJ L 2024/1689, 12 July 2024