Free AI courses worth your time (from the people who built it)

There is a lot of free AI training online. Most of it is either marketing-thinly-disguised-as-curriculum, or hype-cycle content that aged badly six months after publication.

This is a short, opinionated list of free training that actually holds up. All of it is published by the labs and companies that built the technology — Anthropic, OpenAI, Google, Hugging Face — plus one university course that's been quietly running for years and is still the best "what is AI" intro out there.

For each: what it's good for, who it's for, and what to skip.

If you want to learn fundamentals (no coding required)

Elements of AI — University of Helsinki

elementsofai.com

The best non-technical "what is AI" course in existence. Free. 6 weeks at ~2 hours per week, but you can skim it in a weekend. Covers what AI is, what it can and can't do, and the methods behind common applications. Translated into 20+ languages.

Good for: Anyone who wants the conceptual map without the math.
Skip if: You already understand the basics. Move on to a hands-on course.

Google's Machine Learning Crash Course

developers.google.com/machine-learning/crash-course

A 15-hour structured intro that goes deeper than "what is AI" but stops short of "build your own model." Mix of video and interactive exercises. Updated more often than most.

Good for: Engineers, analysts, or PMs who want to understand ML from the inside without going all-in.
Skip if: You want LLM-specific content. This is classical ML — supervised learning, neural networks, classification. Foundational, but not what's on the front page right now.

If you want hands-on with current LLMs

Hugging Face NLP Course

huggingface.co/learn

Free, ~20 hours, ships with working code in their Transformers library. Goes from "what's a token" to "fine-tune a model on your own data." Most up-to-date with what's actually happening in production LLM work.

Good for: Engineers who want to understand how LLMs actually work under the hood, not just how to call an API.
Skip if: You just want to use AI as a user, not understand or modify it.

Anthropic Academy — Claude Code in Action

anthropic.skilljar.com/claude-code-in-action

A free course on agentic coding workflows, taught by the team that built Claude Code. Practical, opinionated, not marketing-y. Covers the patterns that move you from "asking an AI to write a function" to "running an autonomous coding agent on real work."

Good for: Engineers using AI coding assistants who want to graduate past one-prompt-at-a-time.
Skip if: You're not using Claude Code or a similar agentic coding tool. The patterns transfer, but the demos are tool-specific.

If you want role-specific applied prompts

OpenAI Academy — Prompt Packs

academy.openai.com → search "prompt packs"

Curated prompt libraries organized by role: Sales, Customer Success, Product, Engineering, HR, IT, Managers, Executives. Each pack is a list of 20–30 actual prompts you can copy and use immediately. No course structure — these are reference, not learning.

Good for: People who want to skip "what is a prompt" and go straight to "what should I type." Especially useful as a starter library for non-technical roles.
Skip if: You want depth over breadth. Each prompt is short — the value is volume, not nuance. See our /prompts library for the long-form versions.

OpenAI Academy — Skill Labs

academy.openai.com/home/resources → filter "Skill Lab"

Downloadable PDF worksheets that walk through specific tasks (ChatGPT for Excel, ChatGPT for Email, etc.). Short, practical, no fluff.

Good for: Filling in specific gaps. The Excel one is especially good if you've been guessing at spreadsheet formulas.
Skip if: You learn better by doing than by reading. The exercises are small.

If you want data + analytics (the non-AI half of the AI job)

Google Analytics Academy

analytics.google.com/analytics/academy

Free, structured, the gold standard for learning Google Analytics specifically. ~35 hours across multiple courses.

Good for: Anyone whose job involves measuring whether AI features are working. Most AI projects fail at the measurement step, not the building step.
Skip if: You're not going to use Google Analytics. The conceptual material applies generally but the hands-on is GA4-specific.

Google Cloud Skills Boost (Qwiklabs)

cloudskillsboost.google

Hands-on labs for Google Cloud, including their ML and AI services. Many labs are free.

Good for: Engineers planning to deploy production AI on GCP, or anyone wanting hands-on exposure to managed ML services without the credit-card-required friction.
Skip if: You're committed to AWS or Azure. The concepts transfer; the lab time mostly doesn't.

What we deliberately left off

A few categories of "free AI course" we don't recommend, in case you're wondering why they're not on the list:

YouTube playlists by individual creators. Often great, often quickly outdated, hard to evaluate quality without watching. If you have a specific creator you trust, follow them. Otherwise the structured options above are better starting points.
"Become an AI engineer in 30 days" boot camps. Most are funnels for paid programs. The free portion is usually shallow.
"100 ChatGPT prompts that will change your life" lists. Almost universally low-quality. Better to learn the patterns (see our /learn/ai/prompt-craft pillar) than to memorize prompts.

A suggested order if you're starting from zero

Elements of AI — concepts
AINews /learn/ai/get-started — practical foundations
AINews /learn/ai/prompt-craft — how to actually write prompts
Pick one applied direction:
- Engineer: Anthropic Academy → Hugging Face NLP Course
- PM / non-technical: OpenAI Academy prompt packs → AINews /prompts for the deeper versions
- Data / analyst: Google ML Crash Course → Google Analytics Academy
- Operator / founder: AINews /skills/founder → OpenAI Academy executive prompt pack

The list above is more than enough to keep most people busy for a quarter. After that, you'll know which direction you want to go deeper in, and the next-step resources are easier to evaluate.