How does your site compare? Speed benchmarks from 2025

What the numbers measure

The Almanac combines two sources: the HTTP Archive, which tests pages in a lab, and the Chrome UX Report, which records real Chrome visitors. Most figures below come from the field data, per website, in July 2025. A site counts as "good" on a metric when at least 75% of its visits meet the threshold. See the Chrome UX Report and lab and field data.

The thresholds for "good":

All three Core Web Vitals

Mobile keeps improving. Desktop improved by only a point in 2025. The Almanac suggests that better browsers, devices and networks, as well as site optimization, may explain the rise.

The picture changes with popularity. Among mobile sites, 51% of the 1,000 most popular sites had good Core Web Vitals. The share fell to 42% for the next 10,000 and 37% for the next 100,000, then rose to 42% for the next million and 48% for the next ten million. The Almanac reads this as a U shape: the biggest sites invest in speed, middle-sized sites combine more features and third-party scripts with less sustained attention, and the smallest sites are simpler and benefit from platform defaults.

Pages deeper in a site tend to do better. Secondary pages passed more often than home pages, with a 14% lead on desktop and 11% on mobile. The Almanac links this to caching, and to home pages being updated more and holding more dynamic parts.

Loading

Three details from the Almanac's look at the largest element:

• An image is the Largest Contentful Paint element on 85.3% of desktop pages and 76% of mobile pages. Text accounts for 14.4% and 23.7%.
• 57% of those images are JPEG and 26% are PNG. WebP is rising, up 4 points in a year, and AVIF is at 0.7%.
• About 16% to 18% of pages serve that image from a different host than the page. See DNS, TLS and connections.

On priority hints, 17% of mobile pages with an image as the main element set fetchpriority="high", up from 15% a year earlier, and only 2.1% to 2.2% use a preload. About 16% to 17% of pages lazy-load their main image, a habit that delays the thing visitors are waiting for. See how to fix slow LCP and lazy loading.

Interactivity

Mobile INP improved by three points and the gap to desktop narrowed from 23 points to 20. The 1,000 most popular sites went from 53% to 63% good. On mobile, home pages did better than secondary pages, 80% against 69%, which the Almanac links to filters, carousels, form checks and third-party widgets deeper in a visit.

Total Blocking Time, a lab measure, told a different story: its median on mobile rose 58% in a year, from 1,209 ms to 1,916 ms, and at the 90th percentile mobile pages face over 7.5 seconds of blocking. The Almanac offers reasons for the mismatch with INP, including more powerful real-world devices that hide code complexity, and sites that optimize the interactions that dominate INP while still running a lot of background work. See Interaction to Next Paint and Total Blocking Time.

Visual stability

Good CLS has risen each year since 2021, from 62% to 72% on desktop. Home pages shift more than other pages: 71% of desktop home pages were good against 73% of secondary pages, and 79% against 81% on mobile. See Cumulative Layout Shift.

What the numbers say about priorities

• Mobile loading and responsiveness are where most sites lose. Phones trail computers on Largest Contentful Paint, First Contentful Paint, Time to First Byte and INP.
• Layout stability is the easiest to get right, and still one site in five on mobile misses it.
• The main image is usually the bottleneck for loading, so how it is found, prioritized and sized matters.
• Heavy JavaScript is the common thread in blocking time and interaction delay.

Using benchmarks without fooling yourself

• Compare like with like. Use the same device type, the same kind of measure (field or lab), and your own origin or page.
• The Almanac warns that "good" may mean something different for individual sites and their users.
• A good average hides a bad tail. Core Web Vitals judge the 75th percentile, so a quarter of visits can be slower than the threshold. See designing for slow networks.
• The data has a date. These figures are from July 2025.
• Your competitors matter more than the web. Compare with sites your visitors also use.
• Passing is not the goal. Faster pages help visitors whether or not a threshold is crossed.

Check your own page

Run the free audit and compare your mobile numbers with the mobile column above, metric by metric. Then note which of your three are below the share of sites that pass.

Sources

• Performance chapter, Web Almanac 2025: Core Web Vitals pass rates, loading, interactivity and visual stability, the largest element and its priority.
• Defining the Core Web Vitals metric thresholds on web.dev: the thresholds and the 75th percentile.