The Fastest Web Host Passes Core Web Vitals Twice as Often as the Slowest

"Does hosting actually affect Core Web Vitals, or is that just something hosting companies say?" It is one of the most-argued questions in web performance, and the field data settles it. Group the web's sites by the provider hosting them, and the best host passes Core Web Vitals more than twice as often as the worst. Pantheon's sites pass 61.2% of the time. Bluehost's pass 30.4%. Same web, same metrics, same three thresholds. The host is the variable.

That spread has real consequences, because speed is not a back-office detail. A slow server delays the moment your content appears, and Google has found that most mobile visitors abandon a page that takes more than three seconds. Slow loading costs conversions, costs search ranking, since Core Web Vitals are a confirmed ranking signal, and costs the trust a visitor extends to a site that feels quick. A hosting choice that doubles your odds of passing is a business decision, not an infrastructure footnote.

To measure it, we took roughly eleven million real websites, grouped them by the provider serving them, and calculated how often each provider's sites pass Core Web Vitals for real Chrome users. This is an infrastructure study of the hosting layer beneath the site: it is about who serves the bytes, not what publishes them. The ranking has a two-to-one gap at its extremes and a genuine surprise in the middle.

The numbers join two sources. Google's Chrome User Experience Report supplies the real-user field metrics, how fast each origin actually is for the people who load it, and Wappalyzer detection identifies the hosting and infrastructure provider behind each origin. This is field data from real visits rather than a lab score. The study covers roughly eleven million origins across 20 providers, each detected on at least 3,000 measured sites, spanning every hosting category from managed WordPress to budget shared hosting to the modern Jamstack platforms.

A site passes Core Web Vitals only when it is good on all three metrics at once: Largest Contentful Paint under 2.5 seconds for loading, Interaction to Next Paint under 200 milliseconds for responsiveness, and Cumulative Layout Shift under 0.1 for visual stability, all measured at the 75th percentile on mobile. We apply a strict rule where a missing INP reading counts as a fail, which matters for hosts that serve many tiny, low-traffic sites. Because a site can sit behind several layers at once, an origin host plus a CDN, the provider rows are non-exclusive.

Disclosure: PageSpeed Matters has partner relationships with some providers named here, including WP Engine and Cloudflare. This ranking reflects measured Google field data only, not any commercial relationship.

Group sites by host and the pass rate runs from 61% at the top to 30% at the bottom. The top belongs to premium managed WordPress hosts. Pantheon leads at 61% and WP Engine follows at 56%. The premium buys a tuned stack: aggressive page caching, a bundled CDN, optimized PHP and database layers, and infrastructure sized so sites are not fighting neighbors for resources. It shows up directly in the field, where WP Engine's sites post a 96.5% good-INP rate and a median LCP of 1.8 seconds. The managed stack does the boring server-side work that a passing score depends on.

The bottom belongs to budget shared hosting, the mass-market providers most small sites reach for by default. Bluehost sits last at 30%, with DreamHost at 33% and GoDaddy at 47%. Shared hosting packs many sites onto shared resources, and the cost is visible: Bluehost's sites have just a 49.8% good-LCP rate and a 2.6-second median load, the slowest on the board. On entry-level shared hosting, the server itself caps what the rest of your optimization can achieve.

Then the surprise. Vercel and Netlify, the modern Jamstack hosts marketed hard on speed and edge delivery, rank near the bottom at 34% and 33%. That looks like an indictment of their infrastructure, and it is not. Their loading is mid-pack, with good-LCP around 58%, but their responsiveness is weak. Vercel's good-INP is 69.1%, the second-worst on the board. Their edge networks deliver the first byte quickly, and then a large client-side bundle bogs down interactivity.

This is the same story our framework study told. Vercel and Netlify are where JavaScript-heavy Next.js apps live, and those apps pay a hydration tax on INP. The hosts are not slow; the sites they attract are heavy, and the strict all-three pass punishes that heaviness. The pure-play CDNs, meanwhile, land where CDNs should, in the middle: Fastly is the best of them at 49%, Cloudflare at 43% across 4.2 million origins, Akamai at 40%, and CloudFront at 33%. A bundled CDN genuinely helps loading, but it cannot fix a heavy page or slow interactivity, so adding one does not by itself move a site into the passing column.

Read the full board and it measures two things at once, which is what makes it useful. The managed hosts at the top really are running faster infrastructure, and the Jamstack platforms near the bottom really are hosting the heaviest apps on the web. WordPress.com is the clearest example of the strict rule biting: its sites post 87% good-LCP and 95% good-INP, yet its all-three pass reads a middling 40% because it hosts an enormous number of tiny blogs that never register an INP score, which counts as a fail. Nginx and LiteSpeed, web-server software rather than hosts, are shown for context and trend to the web's average.

Nginx and LiteSpeed are web-server software, not hosts, shown for context. Figures are 75th-percentile mobile field values.

If you are picking a host, the spread is the headline: Pantheon's sites pass 61.2% of the time, Bluehost's 30.4%, and the host is the only variable that changed. Premium managed WordPress buys the difference by doing the server work for you, caching, a bundled CDN, and tuned PHP, which is why WP Engine posts a 96.5% good-INP rate and a 1.8-second median load. Budget shared hosting shows the opposite: Bluehost's 2.6-second median and 49.8% good-LCP are the server itself capping what any front-end tuning can reach. If you are on a bottom-tier shared box and struggling, upgrading is the highest-impact move you can make without touching code, and our ultimate infrastructure and hosting guide covers how.

But read the board as two problems, not one. Vercel and Netlify rank near the bottom at 34% and 33% not because their infrastructure is slow, their LCP is mid-pack, but because they host heavy JavaScript apps that fail on INP, Vercel at 69.1% good. A great host running a bloated app still fails, and a lean app on a slow host still fails. If you are on Vercel or Netlify, your bottleneck is the bundle, not the box. For everyone else, the managed hosts at the top have made the infrastructure half easy, and the site owner still has to ship a page worth serving quickly. When both halves line up, passing Core Web Vitals stops being a fight.