# Core Web Vitals 2026: INP, LCP & What Google Now Measures
Core Web Vitals have evolved. Since Google replaced First Input Delay (FID) with Interaction to Next Paint (INP) in 2024, the bar for "good interactivity" has gotten significantly stricter — and a lot of sites that passed CWV under the old rules now fail.
Here's where we stand in 2026.
## The current Core Web Vitals
| Metric | Measures | "Good" threshold |
|--------|----------|------------------|
| **LCP** (Largest Contentful Paint) | Loading | < 2.5s |
| **INP** (Interaction to Next Paint) | Interactivity | < 200ms |
| **CLS** (Cumulative Layout Shift) | Visual stability | < 0.1 |
## What changed with INP
FID only measured the *delay* before the browser started processing the first interaction. It was easy to game. INP measures the *full* time from interaction to the next visual update — and it tracks across the entire session, not just the first click.
This means:
- Heavy JavaScript handlers that block the main thread now hurt
- Long-running React renders show up in real-user metrics
- Sites with chat widgets, analytics scripts, and ad networks suffer
- The "long tail" of interactions matters, not just the first one
## How to fix common INP issues
### 1. Split up long tasks
Anything over 50ms on the main thread is a problem. Break up work with `scheduler.yield()`, `requestIdleCallback`, or `setTimeout(..., 0)`.
### 2. Defer non-critical JS
Third-party scripts (chat, analytics, ads) should load with `async` or `defer`, ideally after user interaction. We use Next.js's `<Script strategy="lazyOnload">` extensively.
### 3. Memoize React components
Re-renders of large component trees on every input change kill INP. `React.memo`, `useMemo`, and `useCallback` aren't optional — they're load-bearing.
### 4. Avoid blocking the main thread on input handlers
Move heavy work off the main thread with Web Workers, or split it across animation frames.
## LCP in 2026
LCP is still 2.5s, but Google now weighs *real-user* LCP from the field, not lab data. Common LCP fixes:
- Serve hero images as next-gen formats (AVIF, WebP) with proper sizing
- Use `fetchpriority="high"` on the LCP element
- Pre-connect to critical origins
- Avoid client-side rendering for above-the-fold content
## CLS — still the easy win
Layout shift is mostly solved by:
- Always specifying image dimensions
- Reserving space for ads and embeds
- Avoiding font swap (use `font-display: optional` or self-hosted)
## Why this matters more in 2026
Beyond rankings, CWV directly impacts conversion. Sites in the "good" range across all three metrics see 20-30% higher conversion than those failing one or more. AI Overviews and generative search results also surface fast sites disproportionately.
We measure CWV on every britRank build with a CI gate — a deploy fails if any metric regresses. [Talk to us about a performance audit.](/services/web-design)