Back to School Digital: EdTech Trends 2025

September starts with a technology homework

The 2025-2026 school year opens with a number that would have sounded like science fiction three years ago: 38% of Spanish schools now use some form of AI in the classroom, per the latest INTEF report (Spain’s National Institute for Educational Technologies). In 2022, it was 4%. That is not gradual adoption; it is a phase shift.

But there is a caveat. “Using AI in the classroom” covers everything from a teacher asking ChatGPT to generate exercises to adaptive tutoring systems that tailor the curriculum student by student. The distance between those two extremes is vast, and most schools are closer to the first than the second. The pattern is not unique to Spain — Ofsted data in the UK and NCES surveys in the US show similar distributions.

AI tutoring: from novelty to curriculum

AI tutoring systems have matured considerably in 2025. Platforms like Khan Academy’s Khanmigo, Duolingo Max, and Photomath (now Google-owned) are no longer experiments: they have millions of active users and measurable efficacy data.

The pattern that works is the Socratic tutor: the system does not give the answer but guides the student with progressive questions. Khanmigo reports a 14% increase in exercise completion rates compared to directly showing the solution. The technical key is that these systems combine a general LLM with a structured curriculum that constrains responses to the pedagogical scope and the student’s level.

For European schools, the barrier is no longer technological but regulatory and budgetary. Per-student license costs range from EUR 3 to 12 per month depending on the platform. With an average ICT budget of EUR 15,000 annually for a 500-student school, the math does not work without external funding.

Adaptive learning: real data at last

Adaptive learning (adjusting content and pace to each student’s individual progress) has been an EdTech promise for a decade. In 2025, it finally has impact data at scale.

A longitudinal study published by RAND Corporation in June 2025 with 12,000 secondary students shows that students in adaptive programs improve by 0.3 standard deviations in mathematics versus the control group. Not a revolution, but statistically significant and reproducible.

Leading platforms (DreamBox, McGraw-Hill’s ALEKS, Century Tech in Europe) share an architectural pattern: a curriculum knowledge graph, a per-student competency model updated in real time, and a recommendation engine that selects the next activity. AI is applied in the recommendation engine, not in content creation. Content is still curated by educators.

This has implications for systems integration. Existing LMS platforms (Moodle, Google Classroom, Microsoft Teams for Education) need to connect with these adaptive tools. The reality is that integrations are fragile: proprietary APIs, incompatible data formats, and roster synchronization that breaks every term.

LMS modernization: the elephant in the room

Moodle dominates the European public education sector: 67% share in Spanish universities, significant presence across France, Germany, and the Nordics. Version 4.3, released in October 2024, adds basic AI features: question generation, automatic forum summaries, and resource suggestions.

But the fundamental problem has not changed. Moodle is a platform from 2002 that has evolved by accumulating features. The user experience is far from what a 16-year-old expects in 2025, accustomed to interfaces like TikTok or Instagram. Canvas and Instructure gain ground in private universities precisely for this reason: they are not necessarily better in functionality, but they are in usability.

The decision is not trivial. Migrating from Moodle to another platform costs between EUR 20,000 and 80,000 for a mid-sized institution (counting data migration, training, and transition period), plus EUR 8,000-15,000 annually for SaaS licensing. Modernizing Moodle with plugins and an updated theme costs between EUR 5,000 and 15,000 but does not solve structural issues. Is there a third option? Possibly — headless LMS architectures that keep Moodle as a backend but replace the frontend entirely are emerging, though still immature.

Public investment: funding is available

Spain’s Kit Digital program, which already funded digitalization for 200,000 SMEs, launched a specific call for educational institutions in March 2025. The allocation is up to EUR 12,000 per school for learning platforms, school-family communication tools, academic management, and cybersecurity.

For public schools, the Ministry of Education’s Digital Education Plan allocates EUR 400 million for 2024-2027, channeled through regional governments. Execution has been uneven: Catalonia, the Basque Country, and Madrid account for 55% of spending to date.

Across Europe, similar patterns emerge. Germany’s DigitalPakt Schule allocated EUR 6.5 billion (though execution delays are well-documented). France’s Strategie du numerique pour l’education commits EUR 1 billion through 2027. The UK’s DfE EdTech strategy is more modest but growing.

For technology consulting firms, this opens concrete opportunities: selection, implementation, and integration support for institutions with budget but limited technical capacity.

Three trends to watch this academic year

AI-assisted assessment. Automated grading of free-response exams (not just multiple choice) is now viable with accuracy above 85% on well-defined rubrics. Gradescope (Turnitin) and Feedback Fruits lead this niche. The time savings for teachers is substantial: from 2 minutes per exam down to 15 seconds of review.

AI-generated content with pedagogical oversight. Tools like Diffit and Curipod let teachers generate activities, presentations, and assessments tailored to their class level in minutes. The teacher provides the learning objective and reviews the output; AI does the heavy lifting of production. It does not replace the teacher. It returns time for direct instruction.

Predictive dropout analytics. Several European universities already run models that predict dropout risk 3-4 weeks in advance based on LMS activity patterns. Early interventions reduce dropout by 8% to 15% according to published data. This may be the AI and machine learning application with the most immediate social impact in the education sector.

What this means for the technology sector

The Spanish EdTech market is worth EUR 1.2 billion in 2025 (AMETIC estimate), growing at 18% annually. Across Europe, the total market exceeds EUR 25 billion. Public funding is available. Demand is real. But the opportunity is not in selling technology to schools that do not know what to do with it. It is in integrating, training, and supporting.

The average school does not need another tool. It needs the five it already has to work together.