Figma's advantage in teaching AI features is that its audience already has design judgement, so the material can skip the fundamentals and get to the point. The help centre articles and courses are well produced, clearly written and pitched correctly for a working designer. That is a better starting position than most vendors have, and the material mostly uses it well. The design system material is the part I would push people towards first.
Features that help maintain consistency, suggest components, name layers sensibly and keep a large file coherent are solving a problem every design team actually has, and the time saved is real and repeatable. It is unglamorous work that nobody enjoys, and automating it is worth more than any generation feature. The documentation covers it properly. Prompt to prototype is genuinely fun and genuinely useful within a narrow band.
Producing a rough interactive prototype from a description in a minute changes what you can do in a stakeholder conversation, because you can explore three directions in the time it used to take to build one. As a sketching tool for ideas you intend to discard, it earns its place. The framing is where I part company. The material presents generated output as further along than it is.
What you get is a plausible arrangement of standard patterns, which is a reasonable place to start thinking and a poor place to stop. It has no opinion about your users, no understanding of your constraints and no reason behind any of its choices, and a designer who ships it has skipped the part where design happens. Experienced designers know this instinctively. Less experienced ones read the marketing and do not.
The handoff problem goes unaddressed and it is the practical failure point. A prototype that looks complete but has no real component structure, no consistent spacing system and no relationship to your codebase creates work for engineers rather than saving it. Guidance on turning generated output into something a development team can actually build from would be the most valuable page in the resource and it does not exist. Plan gating is genuinely confusing.
Which AI features are available on which tier changes, the documentation does not always keep up, and it is common to read about something you cannot access. That is an irritation rather than a flaw and it costs the reader time. Feature coverage dates fast, as it does everywhere in this category. Articles describe interfaces that have moved and workflows that have changed, and the community fills the gap unevenly.
The evergreen design content holds up much better than the AI specific material. Three point five. Competent, well produced material for an audience that can judge the output critically, strongest on the boring consistency work and weakest in implying that generated screens are closer to finished than they are.