SiteGround is one of the most recommended WordPress hosts for speed, and at least one reviewer ranks it first. Yet one independent monitor checked it every 60 seconds for ten months and measured a time to first byte of 779 milliseconds, the slowest result of the seven hosts below.
Both facts are true, because each test times a different stage of loading a page. For a site that needs to rank, the stage that counts most is the one Google measures in real visitors’ browsers.
What Each Speed Test Measured
Most of the evidence in 2026 comes from three independent tests, and they rarely agree on an order.
HostingStep: Server Response Over Ten Months
HostingStep buys real accounts and puts the same test site on each. Then it checks each site every 60 seconds. Its current window runs from November 1, 2025 to August 29, 2026. It records time to first byte from the United States, the response while many test users arrive at once, and a global average from 40 locations. Price counts for 20% of its ranking, so its printed order is not its speed order.
WPShout: Fully Loaded Time
WPShout loads the same small site on each host from six locations. It uses WebPageTest and stops the clock when the last file arrives. More than 29,000 tests have gone into its data since 2021, and the latest update was on September 2, 2026. Fully loaded time is a fair test of a whole page, but it is not a Core Web Vital.
Review Signal: Load Tests by Price Tier
Review Signal is a benchmark series run by Kevin Ohashi. Its 2026 edition, the first in three years, was out by June. It runs load tests on cached and uncached pages, tracks uptime on three monitors, and names a Top Tier in each price band. Hosts choose whether to enter. Of the seven hosts here, only GreenGeeks entered, and it reached Top Tier in both the under-$25 and $25 to $50 bands. A missing score for the other six says nothing about how they would have done.
How Much Host Speed Reaches Rankings
Google says its ranking systems use Core Web Vitals. It also says there is no single signal, and that it will show the most relevant page even when its page experience is weak. Time to first byte is not a Core Web Vital. Still, the browser cannot draw the largest element on a page until the server answers. So a slow server delays Largest Contentful Paint, one of the three Core Web Vitals.
Google’s guidance calls 0.8 seconds or less a good time to first byte. On HostingStep, six of the seven hosts here measured between 338 and 442 milliseconds from the United States, well inside that line. The bigger gaps show up on pages a cache cannot serve, like a cart or a logged-in account. They also show up in how far a page must travel to reach a distant reader.
The Seven Hosts, Test by Test
1. SiteGround
SiteGround lets a site choose data centers in eight countries, including the United Kingdom, Germany, Australia and Singapore, which helps a site whose readers are far from the United States. Every plan also includes server-level caching and a content delivery network, so no caching plugin is needed to get started. Its better independent numbers come from WPBeginner, which measured a Largest Contentful Paint of 608 milliseconds. On WPShout, its 1.31-second fully loaded time placed 11th of 16, ahead of Hostinger and GreenGeeks.
Its weakest figures come from HostingStep, which measured 779 milliseconds to first byte in the United States and 1,200 milliseconds as a global average, the slowest on this list.
2. Kinsta
Kinsta is strongest on the requests a page cache cannot help with. HostingStep recorded a 20-millisecond response under load. It also singled out how well Kinsta handles uncached traffic, the kind a WooCommerce store sends. WebHostingCat measured a Largest Contentful Paint of 342 milliseconds, the fastest in its test, while WPShout placed Kinsta 9th on fully loaded time.
Its sites run on Google Cloud’s C2 and C3D machines over Google’s premium tier network, and a site can be placed in any of 30 data centers. Cached pages go out through Cloudflare Enterprise, with more than 300 edge locations. Plans are metered by visits, so a growing content site moves up a tier.
3. Hostinger
Hostinger runs LiteSpeed servers with caching and a content delivery network on every WordPress plan. On HostingStep, its Business plan answered the load test in 31 milliseconds and measured a 434-millisecond time to first byte. The weaker result is on WPShout, where it placed 14th of 16 on fully loaded time. The entry Premium plan stores sites on standard solid-state drives, while the faster NVMe drives start on the higher plans.
4. GreenGeeks
GreenGeeks markets its plans as fast WordPress hosting, and HostingStep’s figures support part of that, with a 415-millisecond time to first byte and a 26-millisecond response under load on LiteSpeed with LSCache. WPShout’s fully loaded test is the weak result, at 1.56 seconds and 13th place of 16.
5. Rocket.net
Rocket.net serves full cached pages from an edge network it puts at 330 locations, and it placed first on WPShout at 0.40 seconds. WPShout calls it the fastest host in all its data since 2021. HostingStep measured a 338-millisecond time to first byte and a 66-millisecond global average. Rocket.net blocked HostingStep’s load test, though, so there is no independent figure for heavy uncached traffic. Starter allows one install and 50 gigabytes of bandwidth, and every plan includes a web application firewall, malware protection and daily backups.
6. WP Engine
WP Engine has the most consistent record across two sources. On HostingStep it posted 361 milliseconds to first byte, a 27-millisecond response under load and a 65-millisecond global average, and on WPShout it placed 6th. The base Essential tier allows 25,000 visits a month and includes EverCache and a global content delivery network, while automatic front-end optimization costs extra. Higher Essential tiers go up to 30 sites and 400,000 visits.
7. Cloudways
Cloudways lets you pick the server size and the data center yourself, and it bills by the hour. The servers come from five cloud providers across more than 65 data centers. Each one runs NGINX, Varnish and Redis, and the Breeze cache plugin comes already installed. A site can add memory or move closer to its readers without switching hosts. The smallest servers are the problem. HostingStep measured load responses of 272 to 531 milliseconds on them and recommended a 2-gigabyte server to handle load faster. Its WPShout result was 1.02 seconds from the United States, 10th of 16.
When Search Console Says a Host Move Is Worth It
A host move pays off when field data, collected from real visitors, points at the server. Look for these signals before paying for a migration:
- The Core Web Vitals report in Search Console lists page groups as poor or needing improvement on Largest Contentful Paint, and the same templates pass on layout shift and interaction.
- PageSpeed Insights shows a field time to first byte above 0.8 seconds for those pages.
- The failing groups are the uncached ones, such as product, cart, or account pages, while cached articles pass.
- Your visitors are concentrated far from your current data center.
If the report is already green, or the failures are layout shift or slow interaction, the cause is on the page itself, and a new host will not fix it.
