I've bought into six online AI/ML courses over the past three years. Two were worth it. The other four taught me more about marketing copy than about the subject matter — and by the time I figured that out, the refund window had closed.
So I built a five-minute checklist I now run against any syllabus before I pay. None of it requires access to the paid content. All of it is visible on the sales page, if you know where to look.
1. Hands-on ratio, not just "hands-on" as a word
Count the lecture-only modules versus the modules that end in a graded exercise, a project, or a piece of code you write yourself. If the syllabus can't show you a ratio, it's usually because the ratio is bad. I look for at least 40% of listed modules ending in something I produce, not something I watch.
2. A deliverable you can point to afterward
Not "you'll understand transformers" — a deliverable. A working notebook, a deployed model, a portfolio piece. If the course can't name one thing you'll have built by the end, you're paying for a very long video.
3. Last updated date, not just "published" date
AI tooling changes fast enough that a syllabus written 18 months ago is teaching a different landscape. Scroll to the bottom of the page or check the course platform's changelog. If there's no update date anywhere, that's itself information.
4. Refund conditions written in plain language
"Satisfaction guaranteed" is not a refund policy. I want a number of days, a percentage of modules completed as the cutoff, and where to actually submit the request. If I can't find this in under two minutes, I assume it's designed to be hard to find.
5. Who actually taught it, and can I verify that
A named instructor with a searchable history (talks, repos, prior work) is a different risk profile than an anonymous "our expert team." I don't require celebrity credentials — I require a name I can check.
I keep this as a plain-text file, one line per criterion, and literally fill it in before I open my wallet. It takes less time than reading the testimonials section, and it's caught bad purchases the testimonials never would have.
For anyone building out this kind of checklist in Japanese, I found the course-evaluation guide on AI Craft Campus useful as a second reference point — it walks through similar syllabus red flags with local course examples, which helped me sanity-check a couple of my own criteria.
None of this guarantees a good course. It just means you're deciding with the same information the marketing page is trying to get you to skip past.
Top comments (0)