Core Web Vitals 2026: The Complete Guide to INP, LCP and CLS

Two pages go after the same keyword and land next to each other on Google, and both have perfectly decent content. One loads instantly and reacts the second you tap something. The other stutters, jumps around while it's still loading, and takes a beat to respond to a click. A year from now, which one is still sitting in that position? Core Web Vitals are Google's way of measuring that exact gap - not by guessing, but by watching how real visitors actually experience the page.

Core Web Vitals are three measurable signals that describe how a page feels to use in practice: how fast the main content shows up, how quickly it responds once someone interacts with it, and how steady the layout stays while everything is still loading in. They're part of Google's broader page experience signal, sitting next to things like mobile-friendliness and HTTPS - the same basics we've gone over before in why your website isn't ranking on Google.

Largest Contentful Paint (LCP)

LCP is simply how long it takes for the biggest thing on the screen - usually a hero image, a banner, or a large headline - to fully appear. It's answering the question every visitor is silently asking: has this actually loaded, or am I still sitting here waiting?

• Good - 2.5 seconds or less
• Needs improvement - 2.5 to 4.0 seconds
• Poor - above 4.0 seconds

When LCP is slow, it's almost always one of three things: images that were never compressed properly, a server that takes too long to respond in the first place, or CSS and JavaScript sitting in the page's head blocking anything from rendering until they've finished loading. That server-response piece is often the hosting environment underneath the site, not the design, that turns out to be the real bottleneck.

Interaction to Next Paint (INP)

INP took over from First Input Delay (FID) as an official Core Web Vital back in March 2024, and honestly, it's the one most sites are failing right now without even realising it. FID only ever looked at the delay before someone's very first click. INP watches the whole visit - every dropdown, every filter, every add-to-cart tap, not just the first one.

• Good - 200 milliseconds or less
• Needs improvement - 200 to 500 milliseconds
• Poor - above 500 milliseconds

The usual suspects are heavy JavaScript, one-too-many third-party scripts (chat widgets, ad tags, old tracking pixels nobody's audited in years), and event handlers that were never optimised. It shows up hardest on eCommerce sites, where filters, carts, and checkout steps are exactly the kind of interactions INP is built to catch.

Cumulative Layout Shift (CLS)

CLS is about visual stability - how much stuff on the page unexpectedly shifts around while it's loading. It's the metric behind that annoying moment where you go to tap a button and an ad slides in above it at the last second, pushing everything down so you end up tapping the wrong thing entirely.

• Good - 0.1 or less
• Needs improvement - 0.1 to 0.25
• Poor - above 0.25

The fix is usually straightforward: reserve the actual space an image, embed, or ad is going to need before it loads, and avoid dropping new content in above text that's already visible on screen. It's a small detail, but one worth building in from the start rather than patching in afterwards.

Lab Data vs Field Data - Why Your Lighthouse Score Can Lie

Lab data - the kind you get from Lighthouse or PageSpeed Insights - shows how a page behaves in one controlled test, on one simulated connection, at one moment in time. Field data is different. It comes from the Chrome User Experience Report, sits inside Google Search Console, and reflects how the page actually performed for real people, on real phones, over real networks, across the last 28 days. Google ranks using the field data, not the lab score - so it's entirely possible for a page to score 95 in Lighthouse and still be quietly failing Core Web Vitals for the people actually using it, especially on older phones or patchy mobile connections.

The most honest place to check where you actually stand is the Core Web Vitals report inside Search Console itself.

Why This Matters More With AI Search in the Mix

Core Web Vitals won't drag a weak page past a strong one on their own - good content and solid E-E-A-T signals still do most of the heavy lifting. But when two pages are genuinely close in relevance, the faster, steadier one tends to hold its position longer. And as AI Overviews and generative answer engines keep pulling more clicks away from the traditional blue links, the pages that load fast and respond without lag are the ones still getting whatever traffic is left to fight over.

And it's not just about rankings. A checkout page that jumps around or lags every time someone taps it loses sales regardless of where it sits in Google - something we go into in more detail in how a slow website is secretly killing your sales.

A Practical Fix Checklist

• Compress and lazy-load images below the fold; serve them in modern formats like WebP
• Cut or defer scripts you don't actually need anymore - old chat widgets, unused tracking pixels, forgotten ad tags
• Reserve real space for every image, embed, and ad slot before it loads
• Put a CDN in front of the site if visitors are coming from far outside your server's region
• Break big JavaScript tasks into smaller pieces so the browser stays free to respond to clicks
• Self-host and preload the fonts that matter instead of letting them hold up the render

Where This Fits Into a Wider SEO Plan

Core Web Vitals work best as one piece of a bigger technical foundation, alongside good content and solid technical basics. Fixing INP on a page nobody links to internally, or shaving seconds off LCP on content that doesn't actually answer the query, won't move much on its own. Think of Core Web Vitals as the step that makes an already-good page feel as good as it reads - not a fix for pages that aren't good yet.

If you're going page by page through your own site, a sensible order is: get the content and internal linking right first, then run a free site audit to see exactly where LCP, INP, and CLS stand today, and work through whatever it turns up from there.