AI in the classroom is no longer a future question. It is already here, and both students and teachers are using it whether institutions are fully ready or not.
Students are using AI to write, summarise, translate, brainstorm, revise, generate ideas, prepare presentations, check grammar and solve problems. Teachers are using it to plan lessons, create materials, adapt texts, write rubrics, prepare examples and support learners with different needs. In that sense, AI is already part of the learning environment.
But the more we speak with teachers, university staff, trainers and Erasmus+ participants, the clearer one thing becomes: the biggest problem is not simply that students are using AI. The real problem is that the tools have changed dramatically, while many assessment systems have stayed almost the same.
That creates a very difficult situation for educators.
If students can now use AI to produce essays, reports, summaries, presentations and answers, but we still assess them using the same traditional tasks, teachers are left trying to judge old forms of evidence in a completely new reality. They are expected to maintain academic standards, detect misuse, protect fairness, teach responsibly and adapt to a major technological shift, often without enough time, training or flexibility.
This is why the conversation about AI in education needs to move beyond panic, punishment and detection. Academic integrity matters. Students need clear rules. But if our only response to AI is to ask “How do we stop them using it?”, we are missing the bigger issue.
The better question is:
What should learning and assessment look like when AI is part of the world students actually live in?
This is one of the questions we increasingly discuss with Erasmus+ teachers and education professionals who come to Ta Óneira. Many of them already know traditional CLIL methodology. They understand the value of using English to teach content, build communication and create meaningful learning experiences. But now they want to go further. They want to talk about AI, assessment, authenticity, changing classrooms and the future of education.
And they are right to ask those questions.
AI is not just another digital tool
It is tempting to treat AI as one more classroom tool, like a smartboard, a learning platform, an online quiz or a grammar checker. But generative AI changes the situation in a deeper way.
A dictionary helps a student find a word. A search engine helps a student find information. A grammar checker helps a student improve a sentence. AI, however, can help produce the whole answer.
That difference matters.
If a student is asked to write an essay, AI can draft it. If they are asked to summarise a text, AI can do it. If they are asked to prepare arguments for a debate, AI can create them. If they are asked to produce a lesson plan, solve a problem or reflect on a topic, AI can generate a polished response in seconds.
This does not mean the student has learnt nothing. It also does not automatically mean they are cheating. The problem is more complicated than that. A student might use AI as a tutor, a translator, a brainstorming partner, a writing assistant, a shortcut, a ghost writer or a final editor. Each of those uses is different.
That is why assessment becomes so difficult.
When a teacher receives a final product, they now have to ask questions that were much less urgent before. Did the student understand the task? Did they make the decisions? Did they evaluate the information? Did they use AI responsibly? Did they simply accept the answer? Can they explain the work? Can they defend it? Can they apply the same knowledge in a new situation?
If the assessment only looks at the final product, it may no longer give enough evidence of learning.
The real problem is rigidity
From our conversations with Erasmus+ participants, the central issue is not AI itself. It is rigidity.
Many education systems are still heavily built around traditional exams, essays, written assignments, take-home tasks and standardised assessment formats. These formats can still have value, but they were designed for a world in which producing a polished written answer required much more independent effort from the student.
Now the conditions have changed.
When the tools available to students change, but the goals, tasks and assessment methods do not, teachers are placed in an almost impossible position. They are asked to evaluate knowledge using formats that AI can easily support, imitate or even complete. As a result, the classroom can become tense. Teachers feel they need to police every assignment. Students feel suspected. Institutions look for detection tools. Everyone talks about cheating, but not enough people talk about redesigning the learning.
This is the rigidity problem.
If we keep the same targets, the same tasks and the same evidence of learning, while students have access to completely different tools, frustration is inevitable. Teachers will struggle to assess fairly, students will receive mixed messages, and schools and universities will spend more energy controlling AI than teaching students how to use it intelligently.
That is not sustainable.
Education needs to become more flexible, not less rigorous. Those are not opposites. In fact, flexible assessment can be more demanding, because it asks students to show what they understand in richer, more authentic ways.
AI should be embraced in the learning process
Our position is simple: AI should be embraced in the learning process, but assessment needs to go well beyond the final AI-supported product.
Students need to learn how to use AI responsibly. They need to understand what it can do well, where it fails, how it can mislead, how bias can appear, how to check information, how to improve prompts, how to compare outputs and how to decide whether an answer is actually useful.
Pretending AI does not exist will not prepare students for university, work or real life. Banning it completely may be necessary in some specific assessment conditions, but as a general educational strategy it is limited. Students will use AI outside the classroom. Many will use it in their future careers. The question is whether they will learn to use it critically or secretly.
Used well, AI can support learning. It can help students understand a difficult text, generate examples, compare arguments, prepare for a discussion, receive feedback, practise language, explore different perspectives and organise ideas. For students who struggle with language, confidence or structure, it can also make learning more accessible.
But using AI to learn is not the same as submitting AI-generated work as proof of learning.
That distinction is essential.
A student can use AI to explore a topic and still be assessed on their understanding. They can use AI to generate ideas and still be assessed on their judgement. They can use AI to improve language and still be assessed on their ability to explain, defend, adapt and apply what they know.
This is where teachers need training, tools and institutional support.
Assessment must reveal the thinking
Traditional assessment often focuses on the final product: the essay, the report, the presentation, the written answer, the submitted document. In an AI-rich environment, this is not always enough.
Teachers need to see more of the thinking behind the work.
How did the student approach the task? What decisions did they make? What sources did they trust? What did they reject? What did they change? What did they understand better after feedback? How did they use AI, if they used it? Can they explain their choices? Can they respond to questions? Can they transfer the knowledge to a new context?
This does not mean every task needs to become complicated. It means the evidence of learning needs to be stronger.
Some of the most useful assessment approaches are already familiar to many teachers: project-based work, oral defence, debates, presentations, reflective journals, process portfolios, interviews, practical tasks, in-class production, peer explanation, teacher-student conferences and continuous assessment. AI has not invented these methods, but it has made them much more important.
For example, an essay written at home may now be difficult to assess as independent work. But if the student has to submit drafts, explain their process, discuss their sources, reflect on AI use and defend their argument orally, the teacher gets a much clearer picture.
The final product still matters, but it is no longer the only evidence.
Authentic tasks are harder to fake
One of the most interesting conversations we have had with Erasmus+ participants, including university staff and educators working in tourism and hospitality contexts, is how to design tasks where students may use AI, but still need to produce an authentic response.
This is a much better direction than simply trying to create AI-proof homework.
Instead of asking students to complete generic tasks that AI can answer easily, teachers can design tasks that connect learning to context, judgement, personal experience, live communication or real-world application.
For example, instead of asking students to “write an essay about sustainable tourism”, a teacher might ask them to use AI to identify common arguments around sustainable tourism, then compare those arguments with a real local example, a case study, an interview, a site visit or a professional scenario. The student could then present their position and answer questions.
AI can help with the preparation, but it cannot replace the student’s local observation, judgement and oral defence.
Or take a CLIL task. Instead of asking future teachers to “write a CLIL lesson plan”, we could ask them to use AI to generate two possible lesson ideas, critique both, adapt one for their real learners, explain what they changed and teach a short micro-section to the group.
Again, AI is present, but it is not the whole answer.
The learning evidence comes from the human part: selection, adaptation, explanation, reflection and performance.
This is what authentic assessment should do. It should make the student’s thinking visible.
What this means for CLIL teachers
Many teachers who join Erasmus+ training already know the basics of CLIL. They understand that students can learn language through meaningful content, and that English can be used as a vehicle for history, science, tourism, culture, business, social issues or professional communication.
But AI changes both content and language.
In a CLIL classroom, students may use AI to simplify texts, translate vocabulary, generate summaries, prepare arguments, create presentations, plan projects or improve written English. This can be extremely useful, but it also creates new questions for the teacher.
Is the student using English actively, or outsourcing the language work? Are they understanding the content, or just accepting AI output? Is the task developing thinking, or producing a polished result? Can the student explain the topic without the tool? Can they use the language in a live interaction? Can they apply the content to a real situation?
These are not small questions. They go to the heart of CLIL.
If CLIL is about integrating content, communication, cognition and culture, then AI forces teachers to rethink how those elements appear in the task. A student might produce excellent English with AI support, but that does not necessarily mean they have developed communication skills. They might produce a detailed explanation of a concept, but that does not mean they can use the concept in discussion.
So the next stage of CLIL training is not only about designing interesting content-based lessons. It is also about designing tasks where AI can support learning without replacing thinking.
From course discussion to long-term classroom use
One of the best signs that a training course has worked is that the ideas continue after the participant has gone home.
Recently, one Erasmus+ teacher wrote to us asking that his course confirmation include the fact that we had covered AI as part of the programme. He also mentioned that we had discussed classroom rules, including the idea of letting students themselves help decide what consequences should apply when agreed classroom rules are broken.
That detail matters because it shows how connected these topics are. AI in education is not only about technology. It is also about responsibility, classroom culture, student agency and trust.
The same teacher later tried one of the AI activities we had discussed. He asked AI to interview him with ten questions and correct his answers, and it worked very well. He also created his own audio vocabulary trainer and downloaded a BBC podcast to continue practising. What we loved most was that he said the course would continue to support him in his CLIL classes in the long run.
That is exactly the point.
A good Erasmus+ course should not end when the certificate is printed. It should give teachers ideas, tools and confidence that continue to be useful months later, in their real classrooms, with their real students and their real institutional challenges.
Teachers need flexibility, not just policies
Many institutions are trying to respond to AI by writing policies. That is understandable. Teachers and students need clarity.
But policies alone are not enough.
A policy might say that students must not use AI. Another might say that students can use AI only if they declare it. Another might allow AI for brainstorming but not for final writing. These rules may be useful, but they do not solve the classroom problem by themselves.
Teachers need practical training. They need examples. They need task designs. They need assessment models. They need time to discuss real cases. They need permission to adapt. They need institutional support when they redesign assessment.
Because changing assessment is not easy.
It affects workload, grading, fairness, transparency, curriculum planning and student expectations. Teachers cannot be expected to solve all of this alone, in their spare time, while also being told to maintain old targets and old formats.
If education systems want teachers to respond well to AI, they need to give them the tools and the flexibility to assess knowledge differently.
That means moving beyond panic.
And moving towards design.
AI detection is not enough
Many institutions have focused heavily on detection.
Can we detect AI writing? Can we prove a student used ChatGPT? Can we check whether a text is human or machine-generated?
This is understandable, but it is not enough.
AI detection tools can be unreliable. They can create false accusations. They can punish students unfairly. They can also encourage a cat-and-mouse game where students simply learn to hide AI use better.
More importantly, detection does not solve the deeper problem.
Even if you could detect AI perfectly, you would still need to answer much more important questions. What kinds of AI use are acceptable? What kinds are not? How should students learn to use AI ethically? How should assessment change? What evidence of learning do we value? What skills matter now?
Detection may be part of an academic integrity strategy, but it cannot be the whole strategy.
The stronger response is to design assessment where the student’s thinking, process, context and live communication become visible.
What might better assessment look like?
There is no single answer. Different subjects, ages, institutions and goals need different models. Still, several approaches are especially useful in an AI-rich classroom.
Oral defence allows students to explain and justify their work after submission. This is especially useful in universities, project-based learning and advanced secondary education because it helps teachers see whether students truly understand what they have produced.
Process portfolios shift attention from the final product to the learning journey. Students can include ideas, drafts, AI prompts, notes, feedback, revisions, reflections, mistakes and decisions. This helps teachers see learning as a process, not just a polished result.
Project-based assessment asks students to work on real or realistic problems. AI can support the process, but the student still needs to make decisions, collaborate, communicate, adapt and present something meaningful.
Debates and live discussion are particularly powerful because they reveal understanding quickly. If students have researched with AI, they can then use that preparation in a live debate where they must respond, defend, question, clarify and think on their feet.
Reflective journals can help students explain what they learned, how they used tools, what changed in their thinking and what they still do not understand. Reflection is not perfect as assessment, but it is a strong companion to other methods.
In-class production still has a place too. Some tasks can happen in controlled environments, but this does not have to mean only traditional pen-and-paper exams. In-class production can include planning, speaking, problem-solving, group work, practical demonstrations and short written responses.
AI-use transparency may also become increasingly important. Instead of pretending AI does not exist, students can be asked to document how they used it: what prompt they used, what answer they received, what they accepted, what they rejected, what they changed and what they learned.
This turns AI use into an object of reflection, not a hidden offence.
The goal is not to make assessment easier
Changing assessment does not mean lowering standards. In many cases, good authentic assessment can be more demanding.
It is often easier to produce a generic essay than to defend your ideas in a conversation. It is often easier to submit a polished AI-supported text than to explain why you made each decision. It is often easier to memorise an answer than to apply knowledge to a new situation.
The goal is not to make assessment easier.
The goal is to make it more meaningful.
If students are going to live and work in a world where AI is normal, then education should prepare them for that world. They need to learn how to think with AI without surrendering their thinking to AI. They need to use tools without becoming dependent on them. They need to communicate clearly, ethically and authentically.
They need to understand that the value is not only in the answer, but in the judgement behind the answer.
Why this belongs in Erasmus+ teacher training
Erasmus+ is a brilliant space for these conversations.
When teachers and education professionals travel for training, they step outside their normal institutional routines. That distance matters.
In their own workplace, they may be too busy to rethink assessment. There may be pressure, deadlines, meetings, exams, policies and daily problems. But during an Erasmus+ course, there is time to pause.
Time to compare systems.
Time to speak with educators from other countries.
Time to ask uncomfortable questions.
Time to explore practical ideas.
Time to think about what education should become, not only what it has always been.
This is why AI in education fits so naturally into Erasmus+ professional development. It is not just a technical topic. It is a European education topic. It connects language, digital competence, ethics, assessment, inclusion, communication, professional development and institutional change.
For teachers, university staff, NGO workers and trainers, that makes it incredibly relevant.
What we discuss at Ta Óneira
At Ta Óneira, many Erasmus+ participants come to Gozo not only to improve English, but to rethink teaching in a changing world.
AI is now part of that conversation.
Depending on the group, we may explore questions such as: how can teachers use AI without losing the human part of education? How can students use AI responsibly? How can we design tasks where AI is allowed but authentic learning is still visible? How can we assess speaking, thinking, process and judgement? How does AI change CLIL? How can students use English and AI together without simply outsourcing language production? How can teachers create debates, projects and presentations that require real understanding? How can institutions support teachers instead of only giving them restrictions?
These questions do not always have simple answers.
But they are exactly the questions teachers need space to discuss.
And discussing them in English, with other professionals, in a small and personal environment, makes the training even richer.
The human part matters more than ever
There is a strange paradox in all of this.
The more powerful AI becomes, the more important the human part of education becomes.
If AI can generate content, then teachers need to focus more on judgement. If AI can produce language, then students need more opportunities for real communication. If AI can summarise information, then learners need to ask better questions. If AI can create polished answers, then assessment needs to reveal process, understanding and authenticity.
The teacher is not less important.
The teacher is more important.
But the role changes.
Teachers become designers of learning experiences, guides for ethical tool use, assessors of process, facilitators of discussion, and protectors of human communication in a world full of generated text.
That is not a small role. It is a very demanding one.
And teachers deserve training that recognises that.
Further reading
For wider context, see the European Commission guidance on ethical AI and data in teaching and learning, UNESCO’s guidance on generative AI in education and research, and OECD work on AI and education.
Frequently asked questions
How is AI changing the classroom?
AI is changing the classroom by giving students and teachers access to tools that can generate text, ideas, summaries, translations, explanations, images and feedback. This affects homework, assessment, lesson planning, academic integrity and the skills students need to develop.
Should students be allowed to use AI?
Students should learn to use AI responsibly, critically and ethically. AI can be very useful in the learning process, but teachers need clear rules and better assessment designs to make sure students still show real understanding, judgement and authentic communication.
Why is AI a problem for traditional assessment?
Traditional assessment often focuses on the final product, such as an essay, report or written answer. AI can now help produce those final products, which makes it harder to know what the student truly understands. Assessment therefore needs to include process, explanation, oral defence, reflection and practical application.
How can teachers assess students in the age of AI?
Teachers can use project-based assessment, oral defence, debates, presentations, reflective journals, process portfolios, in-class tasks and AI-use declarations. The aim is to make the student’s thinking and learning process visible, not just assess a final polished answer.
Is AI detection enough?
No. AI detection may have a limited role, but it cannot solve the deeper educational problem. Institutions need clear policies, teacher training and redesigned assessment. The focus should move from catching students to designing better evidence of learning.
What does AI mean for CLIL teachers?
AI changes CLIL because it can support both content and language production. CLIL teachers need to design tasks where students use English, think critically, work with content and show authentic understanding, even when AI is part of the learning process.
Can Erasmus+ training help teachers with AI in education?
Yes. Erasmus+ training can give teachers, university staff, NGO workers and trainers time to explore AI in education, compare experiences with other professionals, rethink assessment and develop practical classroom strategies in a supportive environment.
Final thoughts
AI in the classroom is not going away.
We can ban it, fear it, ignore it or chase it with detection tools, but none of those responses is enough.
The better response is to rethink what we teach, how students learn and how we assess knowledge.
If the tools have changed, the goals and evidence of learning must change too. That does not mean abandoning standards. It means making assessment more authentic, more human and more connected to the skills students actually need.
At Ta Óneira, our Erasmus+ courses give teachers, university staff, NGO workers and education professionals space to discuss AI, CLIL, assessment, communication and the future of education in a practical, human way.
Because this is not only about technology.
It is about what education is for.
And that conversation is worth having properly.