A friend of mine paid £14 for a prompt engineering course on Udemy last spring. Six weeks later she had built a small internal tool that drafts client reports for her agency, and nobody has ever asked to see a certificate. A former colleague took a different route: £1,900 for a part-time bootcamp, twelve weeks of Tuesday evenings, and he still copies half his API code from Stack Overflow.
That doesn’t prove cheap AI courses are better than expensive ones. It does show that price and learning outcomes barely talk to each other in this market. Most of what you pay for at the top end is structure, feedback and a person who notices when you’ve gone quiet. Useful things, but not the same as knowledge, which is now close to free.
Cheap Usually Means Narrow, Not Bad
When a course costs £20 instead of £2,000, something has been stripped out. Usually it’s the human element. There’s no instructor reading your work, no cohort to compare notes with, no deadline forcing you to finish week four on a Sunday night.
What survives is content: recorded lessons, notebooks, exercises, a project brief. For a lot of people that’s the whole job. If you can self-direct and you already know how to debug your own mistakes, a £25 course covering the same material as a £1,500 programme will get you to the same place, just with more Googling along the way.
The Four Price Bands You’ll Actually Encounter
- Free to £10. Vendor documentation courses, university open courseware, marketplace courses on sale. Quality is wildly variable but the ceiling is high.
- £10 to £60. Where most decent cheap AI courses sit. Expect 4–10 hours of video, downloadable code, and a single capstone project.
- £60 to £400. Multi-course bundles, specialisations, and short instructor-led workshops. You start paying for structure here.
- £400 and up. Bootcamps, certificates with a brand attached, and anything promising a career change. The premium is for accountability and alumni networks, not secret techniques.
Where the Genuinely Cheap Courses Live
Marketplaces with permanent sales
Udemy courses list at £70 and sell at £15 roughly every other week, so ignore the struck-through price entirely and check when the course was last updated instead. Coursera lets you audit most courses for free, which gives you the lectures and readings without the certificate or graded assignments. That audit route is one of the few genuinely underused options out there, and this guide to free AI training worth your time covers which audit tracks are actually complete.
The tool vendors themselves
Microsoft, Google, Hugging Face and Anthropic all publish free learning paths tied to their own products. They’re promotional by nature, so treat them as product training rather than theory, but the practical value is real. Google’s Generative AI learning path takes about eight hours and costs nothing.
Open courseware and university leftovers
MIT’s OpenCourseWare has full materials for several machine learning modules, including lecture notes and problem sets. The catch is that they’re course materials, not a course, so there’s no hand-holding and some links have rotted. If you want to know how to separate a good free resource from a dead one, there’s a solid breakdown of how to spot a dud before you invest even an hour.
What a Good £25 Course Looks Like
Concretely, here’s what separates a cheap course worth buying from one that wastes a weekend. It runs five to eight hours, not forty. It was updated in the last nine months, because anything teaching AI tooling from 2022 is a history lesson. It ships runnable code you can execute without hunting for missing dependencies. It has at least one project where you build something small but complete, like a document summariser or a classifier trained on your own spreadsheet. And the instructor shows their working when things break, because watching someone debug teaches more than watching someone succeed.
Red Flags That Make a Cheap Course Expensive
A £15 course that eats three weekends and teaches you nothing has cost far more than a £300 course that worked. Watch for these:
- Preview videos that are all slide decks and no screen share.
- Reviews clustered in a single week, which usually means a launch push rather than sustained quality.
- A curriculum listing tools that were renamed or deprecated eighteen months ago.
- Promises about salary or job placement attached to a course under £50.
- No downloadable resources section at all, so every exercise means retyping code from video.
- An instructor with no public footprint anywhere: no GitHub, no writing, no talks.
Twenty Minutes of Vetting Saves Twenty Hours
Before you pay, do three things. Watch the longest free preview at 1.5x speed and judge whether the instructor explains why, not just what. Read the three-star reviews, which are usually the most honest and specific. Then check the course’s last-updated date against the changelog in the tool’s documentation; if the course predates the last major release, skip it.
One more test that costs nothing: search the course title plus “review” and see whether anyone independent has written about it. Courses with no third-party discussion at all are usually thin, and a quick scan of the top AI courses worth your time will show you what the well-documented ones have in common.
Stack Cheap Courses Instead of Buying One Big One
Here’s a route that costs roughly £60 total and covers more ground than most £800 programmes. Spend your first two weeks on a free vendor path to learn the vocabulary and get comfortable with an API. Weeks three to five go to a £15–£20 marketplace course on the specific thing you want to build, whether that’s retrieval-augmented search, fine-tuning, or automation workflows. Weeks six and seven are for building one small project end to end, badly, and then fixing it.
For the last stretch, spend £30–£40 on a short live session rather than another recorded course. A two-hour online AI workshop where someone looks at your code and tells you what’s wrong is worth more than thirty hours of video at this stage, because the gaps in your understanding are now specific and a person can find them fast.
If you’re later tempted to jump to a bigger commitment, go in with numbers rather than hope. Reading up on what an AI bootcamp actually costs and covers will tell you whether the premium buys anything you can’t assemble yourself for under £100. For most people building workplace skills rather than chasing a first job in the field, it doesn’t. The expensive part of learning AI was never the course. It’s the hours you put in between them.