Google shipped a core update in March 2026 that changed how Core Web Vitals get scored, and most site owners still haven’t caught up. The metrics themselves, Largest Contentful Paint, Interaction to Next Paint, and Cumulative Layout Shift, are the same three signals Google has used since 2024. What changed is the scope: vitals are now evaluated across your entire domain rather than page by page, which means a handful of slow legacy templates can quietly suppress rankings on pages that individually pass every threshold.

The pass rate tells the story. Field data from the May 2026 Chrome UX Report shows only around 56% of tracked origins pass all three metrics at once, even though each metric individually has a much higher pass rate on its own. That gap is the whole problem: most sites do not fail because every metric is broken. They fail because one weak metric, almost always INP, drags the other two down with it. This guide covers the current thresholds, the fixes that matter most, and the order to tackle them in.

1. The Three Metrics and the 2026 Thresholds

Google evaluates all three at the 75th percentile of real visits using field data from actual Chrome users, not lab tests. A perfect Lighthouse score means nothing if real users on real networks are seeing something slower.

Metric	What It Measures	Good Threshold (2026)
LCP	How fast the main content loads	Under 2.5s (competitive sites target 2.0s)
INP	How fast the page responds to every interaction, not just the first	Under 200ms (top sites target 150ms)
CLS	How much the layout shifts while loading or interacting	Under 0.1

2. The Audit Order That Actually Works

Fixing vitals out of order wastes engineering time. Image compression will not save a page whose real bottleneck is JavaScript execution. Work through these in sequence.

3. Largest Contentful Paint (LCP)

LCP measures how long it takes the largest visible element, usually a hero image or headline block, to render. In 2026, the majority of LCP failures still trace back to two causes: unoptimized hero images and slow server response times (TTFB). Fixing those two resolves most failures before a single line of application code changes.

Common Cause	Fix	Typical Impact
Unoptimized hero image	Serve WebP/AVIF, size for viewport, preload it	High
Slow server response (TTFB)	Edge caching, faster hosting, reduce redirects	High
Render-blocking CSS/JS	Inline critical CSS, defer the rest	Medium
Client-side rendering the hero	Server-render above-the-fold content	Medium

4. Interaction to Next Paint (INP)

INP replaced FID in 2024 and now carries equal ranking weight alongside LCP and CLS. Unlike FID, which only measured the very first interaction, INP evaluates every click, tap, and keystroke on the page and reports the slowest one. That single change is why INP is the metric most sites are still failing in 2026: a checkout button that stutters once for half a second can drag your score down even if every other interaction is fast. Fixing INP requires rethinking how your code handles events, not just compressing assets.

Common Cause	Fix	Typical Impact
Long JavaScript tasks	Break work into smaller chunks, yield to the main thread	High
Heavy third-party scripts	Audit chat widgets, ad tags, personalization scripts	High
Unnecessary re-renders	Memoize components, debounce input handlers	Medium
Large hydration cost	Partial or streaming hydration on interactive pages	Medium

5. Cumulative Layout Shift (CLS)

CLS penalizes content that jumps around while the page loads or as the user interacts with it. Google’s 2026 refinements are more forgiving of shifts the user could anticipate, like an accordion opening on click, but still penalize unannounced shifts, most commonly ads or embeds that push content down after the fact.

Common Cause	Fix	Typical Impact
Images without dimensions	Always set width/height or aspect-ratio	High
Late-injected ads/embeds	Reserve space with a fixed-size container	High
Web fonts causing reflow	font-display: swap with matched fallback metrics	Medium
Dynamically injected banners	Reserve space before content loads, not after	Medium

6. What the March 2026 Update Actually Changed

Before this update, a slow archive page or an abandoned old template could sit quietly failing vitals without dragging your best content down. That is no longer true. Google now factors performance across your domain, so a cluster of legacy pages can suppress rankings on pages that pass every threshold individually. For agencies and in-house teams alike, this changes the prioritization model: a single unfixed template is no longer a contained problem, it’s a domain-wide one.

7. Using Claude to Prioritize the Fix List

Once you have PageSpeed Insights or CrUX data across your key templates, Claude is a useful triage layer. Export the per-URL field data, paste it in, and ask it to group your worst-performing templates by root cause rather than by page, since the fix for twenty product pages with the same broken carousel script is one fix, not twenty.

For the other half of the SEO stack, tracking whether these fixes actually move rankings, our guide to the Google Kit covers Search Console, GA4, and the rest of the monitoring setup that tells you whether it worked.