The Core Web Vitals story has shifted twice since most blog posts on the topic were last updated. INP replaced FID in March 2024. AI Overviews now factor page experience into who gets cited. And the thresholds for what counts as “good” tightened.
If your last Core Web Vitals review was before 2024, you’re optimising against an outdated rubric. Here’s the current state.
The three metrics that matter in 2026
1. Largest Contentful Paint (LCP) — speed of meaningful content
LCP measures when the largest visible element on the page (usually the hero image or main heading) finishes rendering.
- Good: ≤ 2.5 seconds
- Needs improvement: 2.5 – 4.0 seconds
- Poor: > 4.0 seconds
The most common failures we see in client audits:
- Unoptimised hero images (uncompressed, no WebP/AVIF, no responsive srcset)
- Heavy hero video that delays the image rendering
- Render-blocking JavaScript above the LCP element (chat widgets, GTM tags loaded synchronously)
- Slow server TTFB (time to first byte) — usually a hosting / database problem, not a front-end one
2. Interaction to Next Paint (INP) — responsiveness
INP replaced FID in March 2024. It measures how quickly the page responds to user interactions (clicks, taps, key presses) over the entire visit, not just the first one.
- Good: ≤ 200 milliseconds
- Needs improvement: 200 – 500 ms
- Poor: > 500 ms
INP is harder to fix than the old FID because it tracks every interaction. Common causes of poor INP:
- Heavy JavaScript event handlers (especially on scroll, resize, mouseover)
- Long tasks blocking the main thread for 200+ ms (third-party scripts are usually the culprit)
- Large DOM trees (>1500 elements) causing slow re-renders
- Re-running expensive layout calculations on every interaction
3. Cumulative Layout Shift (CLS) — visual stability
CLS measures how much the page jumps around as it loads. Buttons that move under the user’s finger, text that reflows when ads load, images that push content down — all bad CLS.
- Good: ≤ 0.1
- Needs improvement: 0.1 – 0.25
- Poor: > 0.25
The fix is almost always the same set of tactics:
- Add explicit width and height attributes to every image and embed
- Reserve space for ad slots and dynamic content with min-height
- Avoid inserting content above existing content (cookie banners, top-bar promos)
- Use CSS aspect-ratio for responsive media that doesn’t have known dimensions
What changed for AI search
Google’s AI Overviews and competing answer engines (Perplexity, ChatGPT) now factor page experience into who gets cited. The signal isn’t direct (“this page passes Core Web Vitals”) — it’s inferred via crawl behaviour:
- Slow pages get crawled less often → less fresh content for AI to cite
- Pages with poor INP suggest poor content quality (heavy ad load, intrusive popups)
- High-CLS pages signal low editorial standards
The compounding effect: a page that fails Core Web Vitals doesn’t just rank worse, it gets cited less by AI summaries, which means the brand mind-share opportunity is lost too.
How to actually measure (vs how Google measures)
The number you see in PageSpeed Insights is a synthetic lab test on a single emulated device. The number that affects rankings is field data — real user measurements aggregated by Google over the last 28 days.
Always check the field data when assessing a page:
- PageSpeed Insights → “Discover what your real users are experiencing” section
- Search Console → Core Web Vitals report
- Chrome User Experience Report (CrUX) for granular real-user data
Lab data tells you whether the page CAN be fast. Field data tells you whether real users are actually getting that experience. The gap between them is usually big and informative.
The 80/20 fixes (in priority order)
From auditing 50+ ecommerce + B2B sites, the order of impact:
1. Image optimisation (LCP)
Convert hero images to AVIF or WebP. Add responsive srcset. Compress to ~70% quality (visually identical at this level). One optimisation pass typically takes LCP from 4s to 2.2s.
2. Defer or remove third-party scripts (INP, LCP)
Audit your tag manager. Most sites have 5-15 third-party scripts loading synchronously. Push them to async/defer or remove the ones that aren’t earning their cost.
Common offenders to defer or remove: chat widgets (load on intent, not on pageload), heatmap scripts, A/B testing tools running on every page, social share scripts.
3. Lazy-load below-the-fold images (LCP)
Add loading=”lazy” to every image not in the initial viewport. This is one HTML attribute and it can shave 1-2 seconds off LCP on image-heavy pages.
4. Pre-define dimensions for everything (CLS)
Width and height on images. Aspect-ratio for video embeds. Min-height for ad slots and dynamically-injected content. These are tiny code changes with disproportionate impact on CLS scores.
5. Critical CSS inlining (LCP)
Inline the CSS needed for above-the-fold rendering directly in the HTML head. Defer the rest. Modern build tools (Next.js, Astro, even WordPress with Rocket) automate this.
What you can stop worrying about
Three things people often optimise that don’t actually affect Core Web Vitals scores:
- Total page weight: a 3MB page can pass Core Web Vitals if the critical render path is fast. Don’t obsess over total bytes.
- Number of HTTP requests: with HTTP/2 multiplexing, request count is largely irrelevant. Focus on payload size and render-blocking resources.
- Time to Interactive (TTI): not a Core Web Vital. Useful diagnostic, not a ranking factor.
The bottom line
Core Web Vitals is no longer a separate optimisation track from “is this page good for users”. The metrics measure user experience proxies, and Google has been trending toward weighting them more heavily — both in classical search rankings and in AI-Overview citation decisions.
Run Search Console’s Core Web Vitals report monthly. Identify the URL groups that are failing. Fix in priority order: LCP first (highest correlation with rankings), CLS second (cheapest to fix), INP third (hardest to fix but lowest impact at the margin).
For most scaling brands, three days of focused engineering work — image pipeline + script audit + CLS fixes — moves a site from failing to passing across most pages. The compound effect on rankings + AI citation rates is worth the investment.