A maths teacher I know spent a Sunday last October working through a free AI course. By nine in the evening she had a certificate, fourteen browser tabs open, and no clear idea what to do differently on Monday morning. She is not unusual.
The courses she took were fine. The problem was that nobody told her which ones to skip. So here is the shortlist I would give a colleague: free AI courses for teachers that actually repay the hours, roughly in the order I would work through them, plus the signals that a course is going to eat your evening and give nothing back.
Free AI training for teachers has quietly gotten good
Two years ago, most “AI for teachers” content was a webinar and a slide deck with three stock photos of robots. Now Google, Microsoft, IBM, Code.org and a handful of universities publish genuinely useful self-paced material, and most of it takes between two and six hours rather than forty.
That creates a different problem. There are now hundreds of options, and a fair number are lead magnets wearing a course costume. A free certificate that takes eleven minutes to earn is not training. It is an email capture form with a progress bar.
Start with Elements of AI, even though it isn’t about teaching
The University of Helsinki’s Elements of AI is the one I recommend first, and it confuses people because it never once mentions lesson plans. Six chapters, advertised at around 30 hours, no coding required, available in a long list of languages, and completely free.
What you get is vocabulary. When a vendor says their school chatbot uses a retrieval-augmented model, you will understand what that means for where your students’ data goes. When a colleague claims AI “just copies things it finds online”, you will be able to explain, in plain English, why that is not quite right. That fluency makes every other course on this list easier to judge. You can move much faster than 30 hours by skipping the optional exercises, and nobody checks.
Free AI courses built specifically for classrooms
- Google’s Generative AI for Educators (built with ISTE, around two hours, self-paced). The best short course for practical work: drafting lesson plans, adapting a text for different reading levels, and writing feedback comments that do not all sound identical. You finish with a certificate and, more usefully, a handful of prompts you have already tested.
- Microsoft Learn’s AI for Educators path. Free modules covering AI literacy and using Copilot in a classroom context. Drier than Google’s offering and slightly more technical, which makes it a good fit if your school already runs on Microsoft 365.
- Code.org’s AI 101 for Teachers. A short video series with discussion prompts, ideal for a department meeting where you need eight people on the same page before the debate starts.
- Common Sense Education’s AI courses. The strongest material on ethics, bias and student data of anything free I have seen. If you teach primary or you are the person others ask about safeguarding, start here rather than with the tool-focused courses.
- IBM SkillsBuild. Free, with digital credentials, and broader than teaching. Useful if you are curious about where AI shows up in administration, careers advice and exam logistics.
- aiEDU’s free teacher resources. Not strictly a course, but free lesson materials and short professional development sessions you can run with your own team.
Check these five things before you give up a Saturday
Vetting takes about four minutes and saves entire weekends.
- A published syllabus with named modules, not a bullet list ending in “and much more”.
- A sample lesson you can read without handing over an email address.
- Clear authorship: a named instructor, a university, or a company with a reputation at stake.
- An honest answer to whether the course is free forever or free until module three.
- Something you produce along the way, not just something you watch.
What is missing tells you more than the marketing copy. If the landing page talks about the certificate and never shows the content, the certificate is the product. There is a longer version of this test in this guide to spotting AI training courses that are worth your money, and every point applies when the price is zero.
Three gaps the free courses leave open
Data and consent. Almost no free course tells you whether you are allowed to paste a student’s support plan into a chatbot. That answer lives in your school’s acceptable use policy and your data protection lead, not in a video.
Your subject. A generic example about marketing emails does not translate to a Year 9 chemistry practical. You have to do that translation yourself, which is why the courses are a starting point rather than a solution.
What happens when the tool is confidently wrong. Every course mentions hallucinations. None prepares you for a student citing a fabricated source in an essay, which is the moment AI actually lands in your working week.
Paid courses do not reliably solve these either, and the gap between a good paid course and an expensive one is often just a better sales page. That is the argument in this breakdown of online AI courses, the good ones versus the expensive ones. Free covers more of the ground than most people expect.
Build one artefact per module
The habit that separates teachers who get value from teachers who collect certificates is simple: after every module, make one thing you can use. A bank of eight differentiated question stems. A marking prompt that returns feedback in your school’s format. A parent email draft you would actually send. A multiple-choice quiz on last term’s topic, with the wrong answers you would expect students to pick.
Then use it on real work within 48 hours. Untested prompts feel like progress and behave like clutter.
If the only free tool you have access to is the assistant built into WhatsApp or Instagram, there is a reasonable case for using Meta’s AI as a study partner while you work through the reading. Ask it to explain a concept back to you at the level of a curious 14-year-old, and you will find out quickly whether you understood it.
When you would rather build than watch
Courses teach you about AI. Building something teaches you how it breaks. LearnAI.org is free and project-led, and an hour spent wiring up a simple quiz generator will teach you more about model limits than three hours of video. You will hit your first nonsense output, notice it, and start building the checking habits that matter.
A four-week plan that survives term time
Forty-five minutes, twice a week. That is the whole commitment.
- Week one: Elements of AI, chapters one and two. No notes, just read and answer the exercises.
- Week two: Google’s Generative AI for Educators in a single sitting, then rebuild one existing lesson with what you learned and teach it.
- Week three: Common Sense’s data and ethics material. Write down two questions for your IT lead or designated safeguarding lead and actually send them.
- Week four: Build one reusable tool, write a short note to your department about what worked, and stop signing up for new courses for a term.
Judge the course by what changed on a Tuesday
Pick a task you do every week. Marking thirty short answers, say. Time yourself before you start the course. Then do the same task using what you learned, and time it again. If the second attempt is not meaningfully faster or noticeably better, the course did not stick, and the fix is practice rather than another enrolment.
Two sessions a week will take most teachers through Elements of AI and Google’s educator course inside a month, which is more than the average person manages in a year of bookmarking. Choose one, finish it, and let the results decide whether the next one earns your time.