Since March 2024, INP has officially replaced FID as the responsiveness metric. According to public 2026 reports, 43% of sites still fail the 200 ms "good" threshold — making it the most commonly missed Core Web Vital. Only 47% of sites pass all three vitals.
If in 2025 you could get away with a decent LCP and ignore the rest, in 2026 Google has increased the weight of Core Web Vitals in its ranking algorithm. Let's see what they actually measure, the current thresholds, and how to figure out in 30 minutes whether your site has a problem.
Core Web Vitals 2026: what they are and why SEO cares
Core Web Vitals are three metrics Google uses to measure real-user experience (Real User Monitoring, not lab tests). The official 2026 thresholds:
- LCP (Largest Contentful Paint) — when the largest in-viewport element renders. Good: ≤ 2.5 s. Needs improvement: 2.5-4 s. Poor: > 4 s.
- INP (Interaction to Next Paint) — how fast the site responds to clicks/taps. Good: ≤ 200 ms. Needs improvement: 200-500 ms. Poor: > 500 ms.
- CLS (Cumulative Layout Shift) — how much the layout shifts during load. Good: ≤ 0.1. Needs improvement: 0.1-0.25. Poor: > 0.25.
Google evaluates them per page, on the 75th percentile of real visitors, separately for mobile and desktop. Translation: if 1 out of 4 users has a bad experience, the page is "poor" — even if it looks perfect in the lab.
LCP: cutting time-to-meaningful-paint
In 2026 the LCP element is almost always a hero image or a large text block. The 4 optimizations that actually move the needle:
- Preload the LCP image with
<link rel="preload" as="image" fetchpriority="high">: typically -300 to -600 ms. - AVIF + WebP fallback: 30-60% smaller than equivalent-quality JPEG.
- Edge CDN (Vercel, Cloudflare, Bunny): TTFB drops from 600 ms to 80-120 ms.
- Eliminate font-blocking:
font-display: swapand self-host on first-party domains.
For a typical SaaS marketing site, applying all four together brings LCP from 4-5 s down to 1.5-2 s.
INP: the real bottleneck of 2026
INP measures the delay between any interaction (click, tap, key) and the next painted frame. It's the metric hurting rankings the most today, and the culprit is almost always the same: third-party JavaScript.
What to do:
- Audit your tag managers: every script blocks the main thread and inflates INP. Drop the ones with no measurable value.
- Move tracking scripts to partytown or a service worker where possible.
- For React/Next.js SPAs, use useTransition and startTransition for non-urgent state updates.
- Break long tasks (> 50 ms) into chunks with
scheduler.yield()— supported in Chrome 129+.
CLS: the legacy of ads and fonts
CLS is the easiest to fix but the first to regress when someone drops in a cookie banner or an embed. The golden rules:
- Set width and height on every image and iframe.
- Reserve space (aspect-ratio) for banners and video embeds.
- Load web fonts with size-adjust to avoid shift when the custom font replaces the fallback.
- For cookie banners, use an overlay (position: fixed) instead of pushing content down.
How to measure your site (free, in 30 minutes)
Three complementary tools:
- PageSpeed Insights (pagespeed.web.dev) — gives you the real CrUX (Chrome User Experience Report) data at the 75th percentile.
- Search Console → Page Experience — tells you how many URLs on your domain fail the thresholds.
- Web Vitals Chrome Extension for local debugging during development.
If PSI is green on mobile but red on desktop (or vice versa), Google takes the worse signal: you have to fix both.
Conclusion: priority order
For most SaaS and marketing sites the playbook is: INP first (where you're losing ranking in 2026), then LCP (highest ROI on conversion rate), CLS last (more cosmetic, but it'll regress every time you integrate something new).
On every site we ship we require PSI ≥ 95 on mobile before launch. If your site is slow and you don't know where to start, send us the URL via the quote form: a free first analysis is included. See also our optimisation services for sites already in production.