Core Web Vitals and SEO without a React runtime on content pages

Search and conversion care about what the visitor can read and click — not about which framework hydrated first. Core Web Vitals are an imperfect but useful lens on that reality. This post ties LCP, INP, and CLS-shaped habits to JSkelet’s HTML-first model, without pretending a framework badge substitutes for measurement.

LCP: paint real content early

Largest Contentful Paint rewards pages that show meaningful content quickly. JSkelet returns complete HTML from the server. Titles, prose, and primary CTAs are in the first response. Pair that with the in-memory HTML cache and warm entries spend almost no application time reconstructing the document after the first render.

Practical LCP work in a JSkelet app includes:

• Avoiding hero patterns that wait on client JS to insert the headline.
• Giving featured images honest width/height and predictable aspect ratios.
• Keeping webfonts optional when system stacks are enough — fewer font races on marketing pages.
• Not burying the LCP candidate behind cookie walls or late-discovered carousels.

INP: shrink the JavaScript tax

Interaction to Next Paint suffers when the main thread is busy. Shipping a React/Vue/Svelte runtime to read documentation is a tax you can refuse. On JSkelet, only islands download modules, and only when strategy says so. Theme and nav can be eager; everything else waits for visibility or idle time.

INP-friendly habits:

• Prefer CSS for simple open/close UI when possible.
• Keep island modules small and single-purpose.
• Avoid long tasks in mount — schedule heavy work.
• Do not attach global listeners from every card on a list page.

CLS: reserve space and stop swapping chrome

Cumulative Layout Shift is often a design ops problem: images without dimensions, late fonts, injected banners. Complete HTML helps because structure arrives together, but you can still shift if late islands inject tall UI. Reserve space in templates; treat banners as layout citizens, not afterthoughts.

SEO checklist that stays boring on purpose

• Stable canonical URLs and locale alternates
• Sitemap + RSS for posts
• Semantic headings from Markdown
• Crawlable HTML without executing a client framework
• Fast TTFB from cache hits on public GETs
• Descriptive titles and excerpts written for humans

Boring is good. Rankings rarely reward inventiveness in the <head> at the expense of clarity.

Measure what you claim

JSkelet’s docs and examples emphasise reproducible payload stories for a reason: performance claims should be checkable. Use field data after launch. Lab tools catch regressions in CI; they do not replace CrUX. When you change the island graph or the cache TTL, look at the URLs that matter — home, a docs page, a blog post — not only a synthetic demo shell.

For content marketing, HTML-first + selective islands + HTML cache is a strong default. Add a client framework when the product genuinely needs an application runtime — not because the blog template came from a starter that assumed one.