AI courses vs building: what actually gets you hired.

The signal ladder

From weakest to strongest signal to a UK graduate-scheme recruiter:

1. Weakest
   'Familiar with AI tools' on your CV

   Everyone writes this. It signals nothing.

2. Weak
   Coursera / LinkedIn Learning certificate

   Proves attendance, not capability. Nice-to-have background only.

3. Medium
   Vendor certification (Google, Microsoft, AWS)

   Slightly better — proves you know one product. Still not proof of judgement.

4. Strong
   Personal AI project on GitHub

   Now we're talking. Recruiter can see code, decisions, outcomes.

5. Strongest
   Live tool + demo video + case study

   A clickable URL that opens a working thing solving a real problem, backed by a three-minute video. Beats every other option combined.

The free courses actually worth your time

• Karpathy's LLM series (YouTube). The clearest explanation of how models work. 1 hour, free.
• DeepLearning.AI Prompt Engineering. 1 hour, free. Teaches structured prompting.
• Anthropic Prompt Engineering course. Claude-specific, interactive, free.
• fast.ai Practical Deep Learning. If you want technical depth, this is the one. Free.

Total: about six hours of high-quality background. That's your entire course budget. Everything after that should be building.

Where courses fail graduates

Three reasons the course-first path underperforms for job seekers:

1. No accountability. 90% of paid course buyers never finish.
2. No output. A completed course is invisible to a recruiter until you build something with it.
3. No feedback loop. You don't know if your work is good, and neither does the platform.

A guided, reviewed build fixes all three. That's the entire reason CareerGlowUp exists as the Build and not a course library.

Frequently asked

Are AI courses worth it in 2026?

For learning fundamentals, yes — most are free and short. For getting hired, they're weak signal. Recruiters can't verify what you learned from a Coursera certificate; they can verify a live demo link. Use free courses as background; use a build project as evidence.

What are the best free AI courses?

Three worth your time: (1) Andrej Karpathy's YouTube series on LLMs — the clearest explanation of how the technology actually works. (2) DeepLearning.AI's 'ChatGPT Prompt Engineering for Developers' — free, 1 hour, teaches prompt structure. (3) Anthropic's prompt engineering course — free, teaches Claude specifically. Four hours total. Skip the 40-hour bootcamps.

Which paid AI courses are actually good?

For non-technical graduates, almost none justify their price versus building a real project. If you want structured technical depth, fast.ai's practical deep learning course is free and legendary. For applied AI-at-work skills, spend the money on a guided build programme or a coach instead — the accountability and feedback are what course platforms can't give you.

How do employers actually verify AI skills?

They click your link. That's it. A working tool with a demo video is worth more than any certificate stack because it's self-verifying. Second place: a technical interview where you can walk through decisions. Third: a written case study. Certificates barely register — they've been Amazon-reviewed into meaninglessness.

Should I get an AI certification from Google, Microsoft or AWS?

Only if the job description explicitly asks for it (rare at graduate level). Vendor certs prove you can use one company's product; they don't prove judgement. For a graduate-scheme application, a portfolio project beats a Google AI Essentials badge every time.