Figma AI

What it actually does

Figma AI is a layer of generative features inside Figma rather than a separate product. It does AI wireframing from a text description, text-to-design for UI components, contextual placeholder content, image enhancement and layout suggestions. Figma Make, the prompt-to-app piece, turns a prompt or an existing design file into something clickable with real interactive logic and accessibility labels rather than static frames.

The reason to care is location. It runs where your design files, Dev Mode and Slides already live, so nothing has to be exported and re-imported to get value out of it. For ideation and for getting a rough prototype in front of a stakeholder the same afternoon, it is the fastest path we know of. Beyond that first stretch, it gets shaky, which we cover below.

What it costs

The free tier includes basic AI features with limits. Professional is $12 per seat per month.

The real cost arrived with credits. Figma has been rolling out credit-per-prompt billing from around March 2026, sitting on top of a paid seat you already bought. In practice that is a price rise for anyone who actually uses the AI. Designers report a single average Figma Make prompt eating roughly 200 credits, so a 3,000-credit monthly allowance works out at something like 15 to 30 prompts. Org-level allowances can be gone in hours.

Two things make that worse. Credits are consumed by faulty or unfinished output with no refund for a broken generation, a point Cybernews raised in January 2026 when it noted the shift left "some users worried about paying for faulty or unfinished output". And the accounting feels arbitrary from the user side: people report consumption and results changing mid-month, others appearing to sail past the stated limit, and no visible ledger to check any of it against.

Where it falls down

Figma Make holds up for the first five to ten prompts and then loses the thread on anything with real structure. That is the recurring report, and it lines up with what the tool is good for. It is an ideation aid that is being sold as a path to shipped product.

The code is the other half of the problem. Developers do not recommend deploying Figma Make output to production without rewriting it, as Cybernews put it: "Developers online do not recommend deploying Figma Make's code to production without optimizing it." So the design-to-app promise ends in another handoff, which is the exact handoff it claimed to remove. Designers describe the AI output as unreliable for real UX work, fine for personal projects, not for anything customers will touch.

Figma Make was still labelled Beta while being sold on paid plans. And underneath all of it sits a long-running set of complaints about the core app that the AI push depends on: broken components, lost overrides, lag and glitches on large files.

One more thing to hold in mind. There is no dedicated G2 or Capterra listing for the AI features. Figma the parent product scores 4.7 on G2 across 2,052 reviews and 4.7 on Capterra across 875. Those are ratings of Figma, not of Figma AI, and anyone presenting them as an AI score is misleading you.

Who should buy it

Product and UI designers already paying for Figma should turn it on and use it for early exploration, wireframe starts and quick clickable prototypes pulled from files they already have. At $12 a seat, with credits watched carefully, that is a fair trade.

Do not buy seats for Figma AI on its own merits, and do not budget a sprint around Figma Make producing shippable code, because the reports say you will rewrite it. Teams whose whole goal is prompt-to-production front-end should look elsewhere and keep Figma for design.

If you are not already a Figma shop, the AI features are not a reason to become one. They are a convenience for people who are.