Core Web Vitals — how fast the main content loads, whether the layout jumps around while it does, how quickly the page responds to a tap — are a ranking signal, not just a UX nicety. Vibe-coded sites fail them for boring, fixable reasons. The framework underneath is rarely the problem.
Largest Contentful Paint usually fails on one thing: an uncompressed, unsized hero image. We just ran this exact fix on our own site — a hero image at over a megabyte, converted to WebP, dropped more than 90% in file size with no visible quality loss. That is typical, not exceptional, for a build where nobody thought to compress an image before shipping it.
Cumulative Layout Shift usually fails on the same root cause from a different angle: images without explicit width and height, so the browser doesn't reserve space and the page jumps as they load in. It is an easy miss when a UI was assembled from generated components that never set dimensions, because the prompt that produced them never had to look at the rendered result.
Interaction responsiveness fails when a prompt pulls in an entire library to render one small component, and that unused JavaScript ships to every visitor anyway. The fix is not a rewrite. It is the same production discipline as everything else on this blog: measure real pages with Lighthouse or PageSpeed Insights, fix the top offender, remeasure. Most vibe-coded MVPs never get measured once before launch.
