On March 18, 2026, Google dropped a Stitch update that sent Figma's stock down 8.8% in a single session. The design community reacted immediately: some calling it the "Figma killer", others dismissing it entirely.
I took the time to test it seriously. Not to settle the debate, but to understand what it actually changes in the day-to-day work of a designer delivering products to production.
The short version: Stitch doesn't replace Figma. But it genuinely changes the ideation phase. And that's real.
What Google Stitch Actually Is
Stitch is an AI design tool from Google Labs, built on Gemini. It generates complete UIs from text descriptions, reference images, or vague design intentions. The March 2026 update added three significant things: an infinite canvas (like Figma), real-time voice interaction, and what Google calls "vibe design."
Vibe design is the real paradigm shift. Instead of describing a mockup component by component, you describe an intention like "an interface that builds trust for senior users" or "a dense but readable dashboard for field operators", and Stitch generates multiple design directions that interpret that intention. In 10 to 15 minutes, you have 6 to 8 visual directions to choose from.
Honestly: the first time I saw this work, I understood why markets reacted the way they did.
What Stitch Does Really Well
Exploring directions in the 0→1 phase
This is where Stitch is unbeatable. When a project starts and there's no clear visual direction yet, the classic exploration time in Figma is long: you start from inspiration, do research, lay down some blocks, iterate. Easily 2 to 3 days to produce 3 or 4 presentable directions.
With Stitch, that same exploration work takes 30 to 45 minutes. Not because the quality is equivalent (it isn't yet), but because the volume of material produced lets you make direction decisions much faster.
👉 Ideation that used to take 3 days now happens in a morning. What you do with those directions afterward is where the real design work begins.
Direct connection to code
The other genuine novelty in the March 2026 update: Stitch exposes an MCP server and SDK that connect it directly to development environments like Claude Code, Cursor, or Gemini CLI. In practice, this means designs generated in Stitch produce clean HTML/CSS, directly usable by developers.
On projects where the need is simple and speed to production is critical (landing pages, confirmation screens, error pages), this frictionless design-to-code pipeline is a real acceleration.
Near-zero barrier to entry
Stitch is free. 350 generations per month in standard mode, 50 in experimental mode (the most precise, which accepts reference images). Compared to Figma's roughly $660/month for a 20-person team, the proposition is hard to ignore for a founder who wants to explore a visual direction before investing in a designer.
What Stitch Doesn't Do (and That's Where Everything Lies)
No governed design system
A design system is a set of stabilized decisions: a color palette, typography, reusable components with their variants and states, spacing rules. It's what makes an interface consistent across 40 different screens, maintainable 18 months later, and extensible without breaking everything.
Stitch generates interfaces. It doesn't create systems. Each generation can produce slightly different components with varying conventions. For a production product, this absence of governance is a dealbreaker.
"An AI-generated interface without a design system is a beautiful facade without a foundation. It holds up for the demo, not for the long run."
No simultaneous collaboration
Stitch's infinite canvas lets you manage multiple directions in parallel, which is useful. But it doesn't allow two designers to work on the same file simultaneously, the way Figma does. For a team, that's a concrete daily limitation.
No interactions or animations
Stitch produces static layouts. It handles neither microinteractions, nor transitions, nor dynamic states. Everything related to the experience in motion, often what makes the difference between an interface that's "fine" and one that people actually want to use, remains entirely to be designed in Figma or in code.
Generation limits
350 generations per month in standard mode sounds like a lot. In practice, on an intensive exploration project, you can consume 80 to 100 in two days. Experimental mode, capped at 50, runs out even faster. These constraints aren't blocking for reasonable use, but they need to be anticipated.
The Workflow That Makes Sense
After several weeks integrating Stitch into real projects, here's how we actually use it:
| Phase | Tool | What we do | Estimated time |
|---|---|---|---|
| Exploration | Stitch | Generate 6 to 8 directions, decide on the main path | ½ day |
| Enriched brief | Stitch + Figma | Validate direction with client, lay visual foundations | 1 day |
| Production design | Figma | Build design system, cover all screens, states and variants | The core of the project |
| Dev handoff | Figma + Stitch | Figma specs for developers, Stitch HTML/CSS extracts for simple components | Ongoing |
The simple rule: Stitch to explore, Figma to produce. The two aren't competing: they cover different moments of the same process.
This workflow reduces time spent in the exploration phase by 40 to 60%, depending on project complexity. That recovered time gets reinvested in what actually matters: UX rigor, system consistency, and quality of experience on critical user flows.
What This Means for Founders
If you work with an agency or a designer, Stitch changes something important about how you can prepare your briefs.
Before, describing a visual direction to your design team required either a moodboard, a lot of words, or an approximate reference. Now you can generate 3 directions in 20 minutes to say "it looks more like this". That changes the quality of the starting conversation.
This isn't you doing the designer's work. It's you giving them more material to move faster in the right direction.
👉 Founders who test Stitch before their design meetings arrive with concrete visual intuitions. Briefs become more precise, iterations fewer.
The Real Question: Where Is This Going?
This is where the comparison with AI in development becomes relevant.
Three years ago, GitHub Copilot was supposed to "replace developers." What actually happened: senior developers became twice as productive, repetitive tasks were absorbed by AI, and value shifted toward architecture, technical decisions, and execution quality, exactly what a tool can't do in place of a human.
Stitch follows the same trajectory. It absorbs the part of our work that was time-consuming without being creatively differentiating: generating alternatives, producing material quickly, exploring without investing. It frees up time for what remains fundamentally human in design: understanding users, designing flows that reduce friction, building systems that hold up over time.
In 18 months, Stitch will probably be better at design systems. It may handle states and simple interactions. The boundary between ideation and production will keep shifting.
What won't change: a product you want to use is a product designed by someone who understood its users. AI generates shapes. It doesn't generate understanding.
"The best designers won't be those who resist these tools. They'll be those who use them to do work that AI can't do alone."
In Summary
Stitch is a serious tool, not a gimmick. It genuinely accelerates the ideation phase and opens an interesting design-to-code pipeline for simple components. But it doesn't build design systems, doesn't handle team collaboration, and doesn't replace UX rigor in a production product.
The real move isn't "Stitch vs Figma." It's understanding where each is strong and building a workflow that gets the best out of both.
👉 If you want to understand how these new AI workflows integrate into a real design project, we can talk through it in 30 minutes.