I count 87 courses on our platform right now with "AI" in the title. Twenty months ago there were four. That is not organic pedagogical evolution. That is a market responding to what schools are willing to pay for.
Which is fine. Schools are asking for AI training because their teachers need it. Providers are supplying what schools want. That is how markets work.
The problem is that a lot of those 87 courses are not actually about AI. They are digital-skills courses from 2022 with the word "AI" pasted into the title and one session added on ChatGPT. If you book one of these by mistake, your teachers come back with skills they could have got from a 40-minute YouTube video, and you have burned €500 per person of your grant.
I want to write down how I would evaluate an AI course today, if I were the coordinator making the booking.
The three signals that separate real AI courses from repackaged ones
1. Look at the syllabus, not the title.
The course description will tell you almost everything you need to know. Real AI courses spend at least half their contact hours on hands-on work with actual tools. Not "an introduction to how large language models work." Not "the future of AI in education." Not "ethical considerations." Those topics matter, but as one session out of ten, not five out of ten.
If the syllabus has more than two sessions of pure theory, or more than one session on "the ethics of AI," it is probably a repackaged course. Real practitioners want to show teachers how to actually use the tools.
2. Ask which specific tools they teach.
A good AI course in September 2026 will name specific tools. Claude, ChatGPT, MagicSchool, Curipod, Diffit, Perplexity, NotebookLM, and so on. They should be willing to tell you which tools, in what depth, and why they chose those over alternatives.
Vague answers like "we cover leading AI platforms" are a bad sign. It usually means they picked one general-purpose tool, run through a demo, and call it a course.
3. Ask when the syllabus was last updated.
This one is brutal but useful. In 2026, an AI course whose syllabus has not been substantially updated in the last six months is out of date. The tools change every quarter. GPT-5 rolled out in December 2025, Claude 4 in June 2026, Gemini 3 in August. If the course materials still reference GPT-4 as "the latest" or "the current," the trainer is not keeping up.
Reputable providers will happily tell you their most recent syllabus update. Sketchy ones will deflect.
The specific things to look for in a good AI course
Beyond weeding out the repackaged ones, here is what a genuinely useful AI course covers in autumn 2026:
-
Prompting patterns for lesson planning. Not "here is how to write a prompt" but structured techniques for using AI to generate differentiated versions of the same material, adapt reading levels, and produce formative assessment questions.
-
Detecting AI-generated student work. Practical strategies, not detector software (which does not work reliably). Includes designing assessments that are AI-resistant by nature.
-
Classroom policies teachers can actually implement. Not "here is the EU AI Act." Actual policies you can put in your Year 9 English handbook next term.
-
Marking and feedback with AI. Time-saving techniques that respect data protection rules. Includes what NOT to put into AI tools (student names, identifiable work).
-
Age-appropriate boundaries. Very different AI use in Year 3 vs Year 11. Good courses go deep on this.
-
The hallucination problem, honestly explained. Teachers need to understand why AI confidently makes up citations, quotes, and facts. Then how to work around it.
If a course covers three or more of these, it is worth considering. If it covers most of them, it is worth booking.
What "AI in Education" courses often skip that they should not
Three areas the market consistently underdelivers:
Data protection. Which AI tools comply with GDPR for use with student work. Most teachers assume "if it is free to use, it is fine." That is not correct. This deserves at least one full session and rarely gets it.
Institutional AI policy. How to write a school-level AI policy, not just classroom rules. Coordinators come back from courses with individual teaching techniques but no framework for the whole institution. Then it becomes a headteacher problem that no one has thought about.
Cost realism. Free tools get you to a certain point. The good tools cost money. Good courses talk about what a realistic annual budget for AI-in-classroom tools looks like for a small, medium, and large school. Most courses pretend the whole thing runs on free tiers.
Where the good AI courses are being run right now
Based on what I see through enquiries and provider updates, autumn 2026 has genuinely strong AI-in-classroom courses running in Lisbon, Barcelona and Helsinki. The Nordic providers tend to go deeper on assessment and policy. The Southern European providers tend to go deeper on classroom techniques.
I am not going to name providers here because it will read as promotion. If you send me your topic and dates I can point you to two or three worth shortlisting. Just email and ask.
What to actually do
If AI teacher training is on your Erasmus+ plan for this year, spend a bit more time on provider selection than you would for a more mature topic. Send three enquiries, ask for the syllabus in advance, ask when it was last updated, ask which tools are covered.
The good news: because the field is moving fast, the providers who are keeping up are visibly better than the ones who are not. You can tell them apart in a five-minute email exchange. Take that five minutes.