Critical Thinking was selected because it is the most consistently and strongly affected competence across all partner countries and both data sources. Classroom implementations show that students initially tend to accept AI-generated responses without sufficient questioning, verification, or reflection. However, when structured tasks were introduced, students increasingly evaluated credibility, identified bias, and discussed limitations of AI outputs. Desk research confirms that AI systems can create an illusion of understanding, increasing the risk of cognitive offloading if critical evaluation is not explicitly trained. This makes Critical Thinking essential for preventing passive AI use and for supporting independent judgment in AI-rich learning environments. Across partners, this skill functions as the foundation for responsible, reflective, and epistemically sound learning.

Classroom implementation across Spain, Bulgaria, Türkiye, and Denmark (school partners) consistently shows that students initially tended to accept AI-generated answers without questioning accuracy, completeness, or bias. Observation sheets and teacher feedback document that, at the beginning of the interventions, many students relied on AI for quick answers or idea generation without reflecting on underlying reasoning. However, when structured activities such as source comparison, bias identification, debates, and guided reflection were introduced, students increasingly demonstrated critical evaluation of AI outputs and greater epistemic awareness.

In Türkiye (school partner), Critical Thinking emerged as one of the most frequently observed competences across multiple lessons. Students questioned AI responses, identified missing context, discussed stereotypes and bias, and proposed strategies such as verifying information and thinking independently before consulting AI. The consolidated competence matrix records high frequency and full success for critical thinking, supported by detailed teacher feedback describing a clear shift from trust to critical scrutiny.

Evidence from Bulgaria (school partner) highlights a similar progression. Teachers reported that students initially accepted AI answers at face value, but structured fact-checking and ethical discussion activities led to increased questioning, argumentation, and evaluation of credibility. Critical thinking was particularly visible during debates and dilemma-based activities, where students analysed AI-generated content and reflected on misinformation and societal impact.

Findings from Spain (school partner) show that teachers observed a reduction in students' independent reasoning when AI was used without guidance, reinforcing the need to explicitly foster critical thinking. Teacher feedback stresses that students asked fewer questions and showed less initiative when relying on AI, but that targeted pedagogical strategies helped re-establish analytical reflection and decision-making.

National desk research from Denmark, Spain, and Türkiye (non-school partners) strongly corroborates the classroom evidence. Across these reports, critical thinking is identified as a core competence needed to counter risks such as cognitive offloading, illusion of understanding, bias, and misinformation. The research consistently argues that without explicit focus on critical thinking, AI use may weaken students' independent judgment rather than strengthen learning outcomes. This convergence across research and practice underpins the selection of Critical Thinking as a key competence most impacted by AI exposure.

Problem-solving was selected because AI fundamentally reshapes how students approach and engage with problems in learning situations. Across partner countries, evidence shows that AI can both support and undermine problem-solving processes, depending on pedagogical design. When AI is introduced without constraints, students tend to bypass productive struggle by outsourcing solution paths directly to AI, reducing persistence, strategic thinking, and ownership of the problem-solving process. Conversely, when tasks are explicitly staged and require students to define problems, explore solution strategies, and justify choices before or alongside AI use, problem-solving competence becomes more visible and more robust.

The comparative analysis indicates that problem-solving is particularly sensitive to AI exposure because it sits at the intersection of critical thinking, decision-making, and self-regulation. AI changes not only the speed of reaching a solution, but also the nature of the cognitive work involved. This makes problem-solving a key competence for understanding how AI alters learning processes, especially in tasks that involve inquiry, evaluation of alternatives, and goal-oriented reasoning.

Evidence from teacher notes and classroom observations across Spain, Bulgaria, and Türkiye shows recurring patterns in how AI affects students' problem-solving behaviour. Teachers consistently report that students initially rely on AI to provide ready-made solutions, especially in complex or open-ended tasks. In Spain, teacher feedback highlights that unscaffolded AI use often leads to reduced depth in students' reasoning and a tendency to accept the first plausible solution offered by AI, rather than exploring multiple approaches or reflecting on underlying assumptions.

In Bulgaria, problem-solving became more visible when tasks required collective discussion and justification of solutions. Teachers observed that students engaged more deeply with problems when they were asked to compare their own solution strategies with AI-generated suggestions and to explain why certain solutions were more appropriate than others. This indicates that AI can function as a catalyst for problem-solving when it is used as a comparative reference point rather than as a solution provider.

Classroom observation data from Türkiye provides particularly strong evidence of the conditional nature of problem-solving development. In lessons where AI use was delayed or restricted to later phases, students demonstrated higher levels of independent reasoning, persistence, and strategic exploration of solutions. Observation matrices show that students were more likely to articulate problem definitions, test alternative strategies, and revise their thinking after encountering AI feedback. When AI was introduced too early, however, observations document a clear reduction in students' engagement with the problem itself.

National desk research further supports these findings by framing problem-solving as a competence at risk of erosion through cognitive offloading. Across countries, the literature warns that while AI can increase efficiency, it may weaken students' capacity to engage in sustained problem analysis unless problem-solving is explicitly scaffolded. Taken together, the convergence between classroom data and desk research underlines that problem-solving is not automatically enhanced by AI, but is significantly reshaped by how AI is pedagogically framed, justifying its inclusion as a key skill impacted by AI exposure.

Communication was selected because AI is fundamentally changing how students express, structure, and exchange ideas. Across partner countries, classroom evidence shows that AI tools can both support and weaken communication competence depending on how they are integrated. When AI is used to draft, summarise, or translate text without reflection, students tend to produce generic, impersonal outputs and lose ownership of their own voice. Conversely, when AI is positioned as a feedback partner or a source of alternative phrasing, students become more aware of audience, tone, and clarity, and develop stronger capacity to articulate and defend their ideas.

The comparative analysis indicates that communication is particularly sensitive to AI exposure because it mediates the relationship between thinking and audience. AI changes not only the speed and ease of producing text and speech, but also the authenticity and ownership of the message. This makes communication a key competence for understanding how AI reshapes learning interactions, especially in tasks that involve collaboration, presentation, argumentation, and intercultural exchange.

Classroom implementation across Spain, Bulgaria, Türkiye, and Denmark (school partners) consistently shows that students initially tend to use AI to generate complete written responses without revising or personalising the output. Observation sheets and teacher feedback document that, at the beginning of the interventions, many students copied AI-generated text directly into their work, with limited awareness of audience, purpose, or register. However, when structured activities such as peer review, audience analysis, redrafting, and oral presentation were introduced, students increasingly demonstrated ownership of their message and stronger communicative intent.

In Spain (school partner), teachers reported that students who used AI to draft written assignments often produced text that was grammatically correct but lacked personal voice, structure, and argumentative depth. When students were asked to compare AI-generated drafts with their own writing and to justify their editorial choices, communication competence became more visible. Teachers observed that students developed greater awareness of tone, audience, and purpose, and began to use AI as a tool for revision rather than as a substitute for their own expression.

Evidence from Bulgaria (school partner) highlights the role of oral communication and collaborative dialogue. Teachers observed that when AI was used to prepare talking points or to anticipate counter-arguments, students engaged more confidently in classroom debates and discussions. However, when students relied on AI-generated scripts without rehearsal or adaptation, presentations became flat and impersonal. This indicates that communication competence depends on whether AI is used to support preparation and reflection or to bypass the communicative effort altogether.

Classroom observation data from Türkiye (school partner) provides strong evidence of how AI reshapes collaborative communication. In group tasks where AI was used as a shared resource, students negotiated meaning, compared suggestions, and built on each other's contributions. Teachers noted that AI-mediated collaboration encouraged quieter students to participate and provided common ground for discussion. However, when AI outputs were accepted without discussion, group dialogue diminished and students defaulted to individual, AI-driven work. This pattern underscores that communication competence is not automatically enhanced by AI, but is significantly shaped by how AI is woven into collaborative practice.

National desk research from Denmark, Spain, and Türkiye (non-school partners) strongly corroborates the classroom evidence. Across these reports, communication is identified as a competence at risk of homogenisation and loss of voice when AI is used uncritically. The research consistently argues that without explicit focus on audience, purpose, and ownership, AI use may weaken students' capacity to express original ideas and engage in authentic dialogue. This convergence across research and practice underpins the selection of Communication as a key competence most impacted by AI exposure.

Ethical Judgment was selected because AI introduces new and complex moral questions into primary and lower secondary classrooms. Across partner countries, evidence shows that students regularly encounter issues such as bias in AI outputs, attribution of authorship, privacy, fairness, and the responsible use of AI-generated content. Without explicit ethical guidance, students tend to treat AI outputs as neutral and authoritative, overlooking the values and assumptions embedded in AI systems. When ethical reflection is built into classroom practice, students develop the capacity to question AI outputs, recognise bias, and make reasoned choices about when and how to use AI.

The comparative analysis indicates that ethical judgment is particularly sensitive to AI exposure because AI systems encode and amplify social values, biases, and assumptions that students are rarely equipped to recognise. AI changes not only the content students engage with, but also the moral landscape of the classroom, raising questions about truth, fairness, responsibility, and authorship. This makes ethical judgment a key competence for understanding how AI reshapes the values and norms that govern learning, especially in tasks that involve information use, collaboration, and creative production.

Classroom implementation across Spain, Bulgaria, Türkiye, and Denmark (school partners) consistently shows that students initially tend to treat AI outputs as neutral and authoritative, with limited awareness of bias, attribution, or responsible use. Observation sheets and teacher feedback document that, at the beginning of the interventions, many students copied AI-generated content without questioning its origin, accuracy, or ethical implications. However, when structured activities such as bias identification, source attribution, ethical dilemmas, and class discussion were introduced, students increasingly demonstrated ethical reasoning and responsible decision-making about AI use.

In Türkiye (school partner), ethical judgment emerged as a recurring competence across multiple lessons. Students discussed the fairness of AI-generated content, questioned stereotypes in AI outputs, and debated whether it was acceptable to present AI-generated work as their own. The consolidated competence matrix records high frequency and full success for ethical judgment, supported by detailed teacher feedback describing a clear shift from passive acceptance to ethical scrutiny.

Evidence from Bulgaria (school partner) highlights a similar progression. Teachers reported that students initially accepted AI answers without considering questions of authorship, bias, or fairness, but structured activities such as ethical dilemmas, source attribution, and class discussion led to increased questioning and reflection on the moral dimensions of AI use. Ethical judgment was particularly visible during debates about stereotypes, misinformation, and the responsible use of AI-generated content.

Findings from Spain (school partner) show that teachers observed a tendency among students to present AI-generated work as their own without attribution, reinforcing the need to explicitly foster ethical judgment. Teacher feedback stresses that students rarely raised questions about authorship or fairness when relying on AI, but that targeted pedagogical strategies such as citation exercises and ethical reflection helped re-establish a sense of responsibility and academic integrity.

National desk research from Denmark, Spain, and Türkiye (non-school partners) strongly corroborates the classroom evidence. Across these reports, ethical judgment is identified as a core competence needed to counter risks such as bias, loss of authorship, misinformation, and erosion of responsibility. The research consistently argues that without explicit focus on ethical judgment, AI use may weaken students' capacity to make responsible choices and engage critically with AI-generated content. This convergence across research and practice underpins the selection of Ethical Judgment as a key competence most impacted by AI exposure.

Creativity was selected because AI fundamentally transforms how students generate, refine, and own ideas. Across partner countries, classroom and desk-research evidence demonstrate that AI changes the nature of creative work in primary and lower secondary education. When AI is used to produce finished outputs without reflection, students tend to bypass the generative struggle that underpins creative learning, reducing ownership, originality, and personal expression. Conversely, when AI is positioned as a source of inspiration, a feedback partner, or a tool for exploring alternatives, students develop stronger capacity to generate, evaluate, and refine original ideas.

The comparative analysis indicates that creativity is particularly sensitive to AI exposure because it shows a clear cause–effect relationship between instructional design and learning outcomes. AI changes not only the speed and ease of producing creative work, but also the authenticity, ownership, and developmental value of the creative process. This makes creativity a key competence for understanding how AI reshapes learning, especially in tasks that involve ideation, design, storytelling, and problem-finding.

Classroom implementation across Spain, Bulgaria, Türkiye, and Denmark (school partners) consistently shows that students initially tend to use AI to generate complete creative outputs such as stories, images, and designs without revising or personalising the result. Observation sheets and teacher feedback document that, at the beginning of the interventions, many students accepted AI-generated work as finished products, with limited engagement in ideation, drafting, or refinement. However, when structured activities such as brainstorming, drafting, peer feedback, and revision were introduced, students increasingly demonstrated ownership of their creative process and stronger capacity to generate original ideas.

In Spain (school partner), teachers reported that students who used AI to produce creative work often generated outputs that were technically polished but lacked personal voice, originality, and developmental depth. When students were asked to compare AI-generated drafts with their own ideas and to justify their creative choices, creativity became more visible. Teachers observed that students developed greater awareness of originality, intention, and audience, and began to use AI as a tool for exploration rather than as a substitute for their own imagination.

Evidence from Bulgaria (school partner) highlights the role of structured creative tasks in making creativity visible. Teachers observed that when AI was used to generate starting points or alternative ideas, students engaged more deeply with the creative process, comparing options, refining drafts, and developing personal style. However, when students relied on AI to produce finished work without iteration, creativity diminished and outputs became generic. This indicates that creativity depends on whether AI is used to support exploration and refinement or to bypass the creative effort altogether.

Classroom observation data from Türkiye (school partner) provides strong evidence of how AI reshapes creative learning. In lessons where AI use was delayed or restricted to later phases, students demonstrated higher levels of original ideation, persistence, and personal expression. Observation matrices show that students were more likely to generate multiple ideas, test alternative approaches, and revise their work after encountering AI feedback. When AI was introduced too early, however, observations document a clear reduction in students' engagement with the creative process itself.

National desk research from Denmark, Spain, and Türkiye (non-school partners) strongly corroborates the classroom evidence. Across these reports, creativity is identified as a competence uniquely transformed by AI exposure, with clear cause–effect relationships between instructional design and learning outcomes. The research consistently argues that without explicit focus on ideation, ownership, and refinement, AI use may weaken students' capacity to generate original ideas and engage in authentic creative practice. This convergence across research and practice underpins the selection of Creativity as a key competence most impacted by AI exposure.