Why is the Design Engineer figure so high (80.9%)?
The survey doesn't explain it. A reasonable guess is the role definition: a Design Engineer's output is already code, so AI code generation plugs straight into their work. That is a guess, not a survey finding.
The more interesting question runs the other way: why only 35.0% for IC designers. The survey has no answer there either; it could be task type, or team norms.
Are the people scoring 7.39 satisfied because of vibe coding?
There is no way to tell. The report notes this is a correlation from a cross-sectional survey. Causation could run both ways: people with smooth workflows may be readier to hand work to AI, or AI may be smoothing the workflow, and both stories fit the same numbers.
Also, the survey doesn't give the size of the "nearly all vibe coded" group, so that score may come from a small subgroup.
32.4% of IC designers feel less secure. Should they rush to learn coding?
The survey reports sentiment, not actual career outcomes, so it can't be turned directly into advice. What it does show is that 35.0% of IC designers already put more than half their build time into AI code, so "learning some" is already the norm among peers.
A practical start is to pick one small task you repeat every week, try it for two weeks, see whether the checking cost fits, and then decide how much to invest.
I'm a PM or non-designer. How should I use this data?
Three ways. First, don't treat "50.9% of non-designers already do this" as a reason to follow; it only shows the practice is unremarkable. Second, before starting, separate whether the output is an internal tool, a prototype or a product, since the verification demands differ widely. Third, when someone on the team ships AI-generated code, ask who checked it.
The survey shows 59.1% built a tool with AI in the past six months; many of those are probably the first two categories.
UXtools' Spring 2026 prototyping survey collected 1,478 responses from designers and builders between March 14 and April 6, covering 18 regions, with 61.4% outside North America. By weekly use, Figma leads at 82.6%, with Claude second at 50.8%, then ChatGPT at 48.2%, Claude Code at 38.4% and Figma Make at 34.8%. Five of the top ten tools are AI tools. The part of the survey most worth reading closely is the question about how much build time goes to AI-generated code.
43.8% of respondents say more than half of their build time goes to AI-generated code, while 37.7% do no vibe coding at all. In the last six months, 59.1% built a tool, app or utility with AI; 25.3% do so regularly and another 30.5% want to but haven't. A separate report, State of AI Design 2026, found that 43% of respondents say their companies expect working prototypes as a design output.
Broken out by role, the share spending more than half their build time on AI code is: Design Engineer 80.9%, Lead/Principal 56.8%, non-designer 50.9%, Manager/Director 46.6%, IC designer 35.0%, researcher 26.1%. The researcher group has only 23 people, and the survey itself labels it directional.
The direction of the gap isn't surprising, since a Design Engineer's output is code in the first place. More notable is that non-designers are already past half: PMs and engineers are building interfaces themselves, which echoes State of AI Design's finding that 40% of respondents say PMs and engineers do more design work.
On workflow satisfaction (a 1 to 10 scale), the "nearly all vibe coded" group scores 7.39 and the "no vibe coding" group scores 5.93. It is tempting to read that as "vibe coding makes people more satisfied," but the report itself notes this is a correlation from a cross-sectional survey and does not prove causation. Both directions are plausible: people whose workflow is already smooth may be the ones willing to let AI write most of the code. The survey also does not say how many people are in the "nearly all vibe coded" group.
On how valuable and how secure each role feels, the split is sharp. Among Design Engineers, 50.0% say their role is more valuable and 10.6% say less secure; for IC designers the figures are 24.8% and 32.4%; for researchers, 17.4% and 39.1%. For the next 12 months, the top investment areas are AI-generated coding (64.0%), agent workflows (46.3%) and design systems and tokens (40.2%), while the biggest blocker is the time it takes to learn the tools (55.7%).
If you are a PM or non-designer, 50.9% of your peers already put more than half their build time into AI-written code. The point isn't that you should too. First decide what the output is: an internal tool, a prototype, or a product customers will use. Errors cost little in the first two and need a human check in the third. The survey doesn't separate those categories, so the 7.39 satisfaction score can't be used to prove that production products suit the approach.
If you are spending on behalf of a team, the survey says the biggest sticking point is learning time, not license price. Rather than buying one more seat, a better use of money is to pick one small recurring task, let one person do it for two weeks, record what was saved and what had to be fixed, and then decide whether to scale up.