Figma AI Features in 2026: Everything It Can Do (And What It Still Can't)

Figma shipped more AI features in the last six months than in the previous three years combined. First Draft generates UI from prompts. Make an Image creates and edits images inline. The MCP server connects Figma to Claude Code and Cursor. Figma Make generates full multi-page websites. Some of these are genuinely useful. Some are gimmicks. Here's the honest breakdown from someone who uses Figma daily.

What Actually Works Well

Replace Content is the sleeper hit. It generates realistic text content — names, addresses, product descriptions, dates — to replace lorem ipsum in your prototypes. Stakeholders and testers see something that looks real instead of Latin gibberish. Small feature, genuine time saver.

First Draft generates usable starting-point layouts from text descriptions. You won't ship what it produces directly, but as a starting point that saves 30-60 minutes of frame setup and basic layout work, it's legitimately useful. Think of it as scaffolding, not a finished product.

MCP Server is the most technically significant feature. It connects Figma to external AI agents — Claude Code, Cursor, VS Code. The practical workflow: design in Figma, then ask Claude Code to "read my Figma design and generate production React components." Claude Code pulls your design tokens, component specs, and layout information directly from Figma through MCP. This replaces the manual design-to-code handoff that has been the biggest source of friction in product development for decades.

What Doesn't Work Well (Yet)

Make an Image is competent but not competitive with dedicated image generators like Midjourney, DALL-E, or ChatGPT's image generation. For quick placeholder images within a design, it's fine. For production-quality imagery, use external tools.

Add Interactions auto-generates prototype interactions but often gets the flow wrong. It connects screens in logical-seeming ways that don't match your intended user flow. Faster to set up interactions manually than to correct the AI's assumptions.

Figma Make generates full websites but the output quality varies wildly. Simple marketing sites come out usable. Complex applications produce layouts that need substantial rework. It's a promising direction but not production-ready for anything beyond basic sites.

The Real Story: Figma vs Claude Design

The April 17 launch of Claude Design prompted an existential question for Figma. Anthropic's CPO resigned from Figma's board the same day. Figma's stock dropped 7.28%.

The honest assessment: Figma and Claude Design serve different stages of the design process. Claude Design is faster for going from zero to first prototype. Figma is better for everything after that — collaboration, refinement, component systems, developer handoff. Smart teams will use both.

For a detailed comparison, see our Claude Design vs Figma breakdown. To find the right AI model for your design workflow, try our AI Model Picker Quiz.

What This Means For Designers

AI isn't replacing designers. It's replacing the parts of design that were always tedious — layout scaffolding, placeholder content, basic prototyping, design-to-code handoff. The parts that require taste, judgment, user research, and creative vision remain irreplaceable.

The designers who thrive in 2026 are the ones who use these tools to skip the boring parts and spend more time on the parts that actually require human judgment. The designers at risk are the ones whose value proposition was primarily execution speed on tasks AI now handles in seconds.

Disclosure: Some links in this article are affiliate links. We only recommend tools we've personally tested and use regularly. See our full disclosure policy.