The Median Website Now Loads in 2.0 Seconds, Down From 2.7 Six Years Ago

It is easy to believe the web keeps getting heavier and slower: more scripts, more trackers, more bloat with every redesign. Six years of real-user data say the opposite is happening. The median mobile page now loads in two seconds flat, taps answer twice as fast as they did in 2022, and nearly nine sites in ten hold still as they load. On every measure that matters to a visitor, the web has measurably sped up.

That progress is not academic. Faster pages keep visitors who would otherwise leave, convert more of the ones who stay, and rank better in a search engine that has treated speed as a signal since 2021. When the typical site crosses from failing to passing on loading, millions of businesses gain back the customers a slow first impression used to cost them. The direction of travel is the story, and the direction is up.

We tracked Core Web Vitals across six years and up to eleven million sites a year using Google's Chrome User Experience Report, the real-user field data collected from Chrome visitors across the web. Every one of the three vitals moved the right way, every year. Here is how far the web has come, and where the last frontier remains.

The figures come from the Chrome User Experience Report, Google's record of how real people experience the web, using one release per year from June 2020 through 2025 plus the May 2026 release. For each origin we take its 75th-percentile loading (LCP), responsiveness (INP), and visual stability (CLS), then report the median across all sites and the share meeting each good threshold: LCP at 2.5 seconds or less, INP at 200 ms or less, and CLS at 0.10 or less. Between 7.0 and 11.1 million origins are measured each year. Responsiveness begins in 2022, because before then CrUX reported the older First Input Delay metric.

Take the median website, the one in the middle, and follow it across six years. In June 2020, its Largest Contentful Paint on mobile was 2.7 seconds, over the good line and in the slow majority. By May 2026 it is 2.0 seconds and comfortably passing, about 26% faster. The share of sites with good loading crossed from a minority (46%) to a majority (66%) over the same window.

That single number captures the whole trajectory. The middle of the web moved from failing to passing on loading, and it did so steadily, year after year, rather than in one lucky jump. The typical site's loading time crossed Google's 2.5-second good line around 2022 and has kept falling since.

The other two vitals moved even more decisively. Layout stability is the win nobody talks about. In 2020, only two-thirds of sites held still as they loaded, and janky, jumping pages were normal. By 2026, 87% are stable and the median site's CLS is effectively zero. Most of that gain was banked by 2022, because the fixes were concrete and mechanical: put width and height on images, reserve space for ads and embeds, and load fonts without a reflow. The web learned that lesson first.

Responsiveness improved fastest of all. Measured only since 2022, the median INP dropped from 175 ms to 100 ms, and good-INP rose from 61% to 83%. The timing is telling: the biggest single-year jump came in 2022 to 2023, ahead of Google's promotion of INP to a full Core Web Vital in March 2024. The web optimizes ahead of what it is about to be graded on, and the responsiveness curve is the cleanest proof.

What is left is loading. LCP improved the least in relative terms and remains the metric roughly a third of the web still fails. That makes sense, because it is the hardest of the three, tangled up with server response time, image weight, render-blocking CSS and JavaScript, and the realities of the networks people actually browse on.

If there is a next frontier for the web's speed, it is the same one it has always been: getting the main content painted quickly. Stability is largely solved, and if your CLS is bad in 2026 you are in a shrinking minority with a well-known fix. Responsiveness is mostly good but has a long tail of JavaScript-heavy sites. Loading is the open problem, the metric most likely to be the reason a site fails today, and the one worth auditing first. The overall picture holds when you split it by device too, as the mobile-desktop gap over time shows.

Mobile, median of per-site 75th-percentile values plus good-rates, Chrome UX Report. INP blank pre-2022.

The six-year line points one way on all three vitals, but it does not point evenly. Stability was banked early, climbing from 66% good in 2020 to 87% by 2026 with most of the gain in by 2022, and responsiveness followed, from 61% good INP in 2022 to 83% now with its steepest jump in 2022 to 2023, ahead of the March 2024 ranking change. Loading moved least in relative terms, which is why LCP is still the metric roughly a third of the web fails.

That tells you where to spend attention. If your site fails a vital in 2026, loading is the likeliest reason, because CLS and INP are largely handled by modern defaults while LCP still rewards deliberate work: prioritise the largest image or heading, cut render-blocking CSS and JavaScript, and speed up the server response. The median site only crossed the 2.5-second good line around 2022, so a slow LCP today is common ground rather than a lost cause.

The broader lesson is that the web improves what it is graded on. Every vital moved the right way every year across 7 to 11 million origins once Core Web Vitals became a shared scoreboard and a ranking signal, and the median mobile page fell from 2.7 seconds to 2.0. The sites that keep pace will treat loading, the last open metric, with the seriousness the web already brought to holding still and responding fast.