The state of Astro and WordPress, 2026.

A report for people who run a WordPress website and want to know whether, and how, Astro fits into its future. Statistical figures below were collected from the cited sources on 22 September 2026. Where a figure is a vendor’s own claim, we say so.

Editorial update, 26 September: clarified the limits of the performance and crawler comparisons, added current Google AI reporting controls, and connected the findings to planning resources. The statistical research cutoff remains 22 September; this is not a new measurement of the datasets.

The short version

• WordPress is shrinking faster than it ever has, and it is still the biggest thing on the web. It runs 40.2% of all websites and 58.8% of sites with a known content management system. It lost 0.6 points of share across all of 2025 and 2.8 points in the first nine months of 2026.
• WordPress 7.0 and 7.1 improved the editor and the admin, not the public website. AI tools in the dashboard, an Abilities API for automation, notes, responsive controls. Real-time collaboration was pulled twelve days before 7.0 shipped. Seven in ten WordPress sites already run 7.x.
• Astro is now owned by Cloudflare and is still open source. Downloads were up 6.85 times year on year in August. It has led State of JS in developer retention for three straight years. Astro 7, released in June, added the caching controls that let a live WordPress backend sit behind a fast front end without rebuilding the site on every edit.
• In the cited field dataset, Astro sites pass Google’s Core Web Vitals on mobile at 71%. WordPress sites pass at 48.7%. These are different populations of sites, not a controlled migration experiment or a prediction for your website.
• Readable HTML helps retrieval, and Google says you need no special files for AI search. Crawler capabilities differ and change; do not apply one historical observation to every AI system. In the referral studies discussed below, AI assistants send a small fraction of traffic and Google remains much larger.
• EmDash, Cloudflare’s WordPress replacement, is real and unfinished. It is at version 0.38, still labelled beta, and now runs Cloudflare’s own blog. It imports content from WordPress, not plugins or themes.
• The connector sources reviewed for this report leave substantial integration work with the developer. The reviewed community package was archived, and we did not find the promised Faust.js Astro support by the research cutoff. The built-in WordPress REST API is an available transport, subject to the site’s access rules and content exposure; it does not connect the entire publishing workflow by itself.
• The public write-ups reviewed here include both leaving WordPress and retaining it for editors. They illustrate different choices, but this sample cannot establish which pattern is more common across all migrations.

WordPress in 2026: still the biggest, shrinking faster

WordPress runs more of the web than anything else, and it is losing share faster than it ever has.

On 22 September 2026, W3Techs counted WordPress on 40.2% of all websites, and on 58.8% of the websites whose content management system it can identify. The peak was 43.6% on 1 January 2025. WordPress gave up 0.6 points of share across 2025. It has given up 2.8 points so far in 2026. W3Techs: WordPress usage, W3Techs: content management history

Two cautions before reading too much into that. W3Techs measures all websites, so a site that closes, moves to Shopify, or becomes a single static page all count the same way. The figures say WordPress is shrinking, not where the sites go. And 40% of the web is still an enormous installed base. Yoast, Elementor, Rank Math, and Advanced Custom Fields each report between two and ten million active installs. WordPress.org plugin directory

What WordPress 7.0 and 7.1 actually shipped

WordPress 7.0 “Armstrong” arrived on 20 May 2026, six weeks after its original 9 April date. Its release post leads with AI: “A new AI Client in Core lets WordPress communicate with generative AI models, while connections are easily managed from a single hub in the dashboard.” The Abilities API, a registry of what a site can do that plugins and tools can call, ships alongside it, with a JavaScript counterpart and a command palette in the admin. WordPress 7.0 release post

The headline feature that did not ship is real-time collaborative editing. The core team removed it twelve days before release, citing “surface area, race conditions, server load, memory efficiency, and recurring bugs found through fuzz testing.” 7.0 shipped notes and visual revisions instead. Real-time collaboration removed from 7.0

WordPress 7.1 “Mary Lou” followed on 19 August 2026, released live on stage at WordCamp US in Phoenix. It is an editor release: responsive styles per breakpoint, the admin bar inside every editor, an image crop workflow, inline notes with mentions, faster media handling, and new Playlist and Tabs blocks. No collaboration, no new AI features. WordPress 7.1 release post

Adoption has been quick. By 22 September, 56.3% of WordPress sites reporting to WordPress.org were on 7.1 and another 12.5% on 7.0. The 7.1.2 release on the same day fixed a vulnerability WordPress rated critical. WordPress.org statistics

WordPress 7.2 is scheduled for 9 December 2026, timed to State of the Word, with Matt Mullenweg as release lead. WordPress 7.2 schedule

What this means if you run a WordPress site

Read the last two releases together and the direction is clear. WordPress core is investing in the editing and administration experience: AI in the admin, abilities that let tools operate the site, notes, media, and responsive controls in the block editor. None of it changes how the public website is built or delivered. The public pages still come from a theme, a page builder, or a front end that reads WordPress content from outside.

That split is the story of this report. The editing side of WordPress is getting better. The delivery side is where teams increasingly look elsewhere.

One piece of background belongs here. WP Engine’s lawsuit against Automattic and Matt Mullenweg is in discovery, with a ten-day jury trial scheduled for October 2027. It matters to ordinary site owners for two reasons: WP Engine maintains Advanced Custom Fields and the Faust.js headless toolkit, and Automattic runs WordPress.org, where plugins are distributed. Nothing about the case blocks a rebuild, but it is a reason not to build a site’s future on tools whose owners are in court with each other. The Repository, 11 Aug 2026

Astro in 2026: owned by Cloudflare, still open, and built for live content

The acquisition

Cloudflare acquired The Astro Technology Company on 16 January 2026. Terms were not disclosed. The press release commits that “Astro will remain open source” and that it “will continue to be the best way for developers to build content-driven websites, whether they host on Cloudflare or elsewhere.” Astro’s co-founder Fred Schott wrote the same day that “Astro will remain free, open-source, and MIT-licensed” and that the team “will continue to support and improve deployments for all targets,” not only Cloudflare. Cloudflare press release, Astro: Joining Cloudflare

Eight months on, both promises have held. Astro 6 and 7 both shipped with support for Netlify, Vercel, Node, and Cloudflare. In August the project’s stewardship passed from Fred Schott to Matthew Phillips, a long-time core maintainer, which reads as continuity rather than change. What’s new in Astro, August 2026

For a business, the practical meaning is that the framework has a funded owner with a direct interest in it working well, and a written commitment that you can host it anywhere. Cloudflare’s own interest is obvious. It sells the hosting. That is a reason to check the escape hatches, and they are there.

The numbers

Astro’s npm package was downloaded 20.1 million times in August 2026, up from 2.9 million in August 2025. That is a 6.85 times increase in a year. Raw download counts across the whole ecosystem are inflated this year, most likely by automated builds and AI coding agents, so the fairer signal is relative share. Astro’s downloads as a share of Next.js downloads went from 2.6% in August 2023 to 4.1%, 4.7%, and 8.9% in each following August, and 9.7% in the thirty days to 21 September. npm downloads API

Developer sentiment has been consistent for three years. In the State of JS survey’s meta-frameworks category, Astro’s retention, the share of people who have used it and would use it again, was 90% in 2023 and 94% in both 2024 and 2025, first place each year. Usage among respondents rose from 19% to 27% over the same period, second only to Next.js. State of JS 2025: meta-frameworks

Site counts are harder. Wappalyzer tracks about 360,000 live websites on Astro, 67% of them in the United States. HTTP Archive’s August 2026 crawl found 43,850 Astro origins with enough Chrome traffic to measure, against 2.75 million for WordPress. Whatever the true count, Astro is roughly one to two percent of the size of WordPress by number of sites. It is growing fast from a small base. Wappalyzer: Astro, HTTP Archive Core Web Vitals Technology Report

What Astro 7 changed for WordPress teams

Astro’s history is a series of additions that make it more suitable for content that changes. Content collections arrived in 2023. Live content collections, which read from an external source at request time instead of at build time, became stable in Astro 6 on 10 March 2026. Astro 6

Astro 7, released on 22 June 2026, added the missing piece: route caching. A page can now be cached and tagged, and the site can invalidate exactly the pages that changed by tag or by path. Live loaders return cache hints so the cache knows what a page depends on. All of that is stable in 7.0. The integrations that push those invalidations to a CDN’s edge cache, including Cloudflare’s, are still marked experimental in the documentation as of September. Astro 7, Astro docs: caching

Why this matters for a WordPress site: before this, the honest choices were to rebuild the whole Astro site whenever an editor published, which gets slow for a site with thousands of pages, or to render every page on every request, which gives up much of the speed. Now an editor publishes in WordPress, WordPress tells the Astro site which pages changed, and only those pages are refreshed. That is how a content team keeps working the way it does today while the public site behaves like a static one. The core of it is stable. The last mile to a particular CDN is still settling, and a rebuild project should plan for that.

Astro is at version 7.3.3 as of 16 September. There is no Astro 8 announcement. npm: astro

Performance in the field

The most useful performance comparison is not a benchmark but what real sites achieve. HTTP Archive measures every origin with enough Chrome traffic against Google’s Core Web Vitals thresholds, by technology. For the August 2026 dataset, on mobile:

Source: HTTP Archive Core Web Vitals Technology Report, dataset dated 1 August 2026.

The dataset also reports 90.8% of WordPress origins with good Interaction to Next Paint, and 23.5% with good Time to First Byte on mobile, against 64.2% of Astro origins for the latter measure. That identifies server response as a useful diagnostic question. It does not establish the cause of a particular site’s delay or attribute the Core Web Vitals gap to one metric. Cached WordPress pages need not be assembled from a database for every visit.

The populations differ in size, implementation, hosting, content and functionality. This comparison does not isolate the effect of the framework. Check the actual site’s field data, cache behavior and server response before choosing between optimization and a rebuild. A fair before-and-after experiment must preserve comparable content and functionality; a framework change alone does not guarantee passing Core Web Vitals.

EmDash: the replacement path from Astro’s owner

On 1 April 2026, Cloudflare announced EmDash, a content management system written in TypeScript and built on Astro. The announcement’s own words: “Our name for this new CMS is EmDash. We think of it as the spiritual successor to WordPress.” Cloudflare blog: Introducing EmDash

That makes it the most important thing to understand about Astro’s relationship with WordPress in 2026. The company that owns Astro is not building a bridge to WordPress. It is building a replacement, with an importer.

Where it stands. EmDash is at version 0.38.0 as of 15 September 2026, after roughly weekly releases since June. Its README still says “beta preview.” The 1.0 milestone on GitHub has two open issues left, and Cloudflare has referred to “the upcoming launch to v1” without a date. The repository has 12,447 stars and the package was downloaded about 115,000 times in the last thirty days. EmDash releases

What it imports. EmDash reads a standard WordPress export file, or connects to a WordPress site through its own exporter plugin, and brings across posts, pages, media, and taxonomies. It does not import plugins, themes, or page-builder layouts, and it cannot import directly from WordPress.com. Anything a plugin did on the old site has to be rebuilt. EmDash docs: content import

The lock-in question. On 2 April, Matt Mullenweg wrote that “the plugin security only works on Cloudflare” and that EmDash “was created to sell more Cloudflare services.” Cloudflare’s chief executive called it “a fair critique.” It was fair in April. By late June, EmDash had shipped a sandbox that runs plugins on plain Node.js servers, and the maintainers closed the issue. The criticism is now out of date, though the Cloudflare path still requires a paid Workers plan for the sandbox. Mullenweg: EmDash feedback, EmDash docs: Node.js deployment

Who is using it. Cloudflare moved its own blog to EmDash on 12 August 2026 and published the operating numbers: about 75 requests a second normally, spikes above 5,000, and a 28,000 request-per-second attack absorbed without incident. Beyond that, there are no published site counts, no hosting product, and no showcase. An official plugin registry went live in mid-September with a small catalogue. CMSWire’s verdict from launch week, “right architecture, empty ecosystem,” still describes it. Cloudflare blog on EmDash, CMSWire

Our reading. EmDash is a serious project by serious people, several of them from Astro’s core team, and it will reach 1.0. For a business with a WordPress site in 2026 it is not a destination yet. It has no ecosystem to replace the plugins a typical site depends on, and its own documentation describes it as a beta. It is worth watching because it shows where the Astro ecosystem’s centre of gravity is heading, and because a site that has already moved its public front end to Astro would find EmDash a much shorter step later than a site still rendered by a WordPress theme. Keeping WordPress as the editor today does not close that door. It keeps it open.

What AI systems and search need from a website in 2026

A lot has been written this year about making websites “AI ready.” Most of it is guesswork. Here is what the primary sources actually say.

Check what each crawler can retrieve

Vercel’s December 2024 network study observed that the major AI crawlers it examined did not render JavaScript. That is a dated observation about those crawlers, not a universal statement about current search systems or browser agents. Google can process JavaScript, subject to its rendering and access requirements. Check the retrieval behavior relevant to your audience instead of assuming every AI product uses the same crawler. Vercel: The rise of the AI crawler, Google JavaScript SEO guidance

Putting essential content and links in the HTML response reduces reliance on a crawler running browser scripts. It does not override robots rules, authentication, CDN challenges or search selection. Both conventional WordPress and Astro can deliver useful HTML. Inspect the actual page and edge access before recommending a framework change. Our AI-readiness explanation describes those distinctions; the review service sets out the deliverable and its limits.

Google’s own guidance

Google published a guide to its generative AI search features and last updated it on 10 July 2026. It is short and worth quoting because it contradicts a good deal of what is being sold:

• “You don’t need to create new machine readable files, AI text files, markup, or Markdown to appear in Google Search (including its generative AI capabilities), as Google Search itself doesn’t use them.”
• On the popular llms.txt file: “Doing so will neither harm nor help your site’s visibility or rankings in Google Search, as Google Search ignores them.”
• “Structured data isn’t required for generative AI search, and there’s no special schema.org markup you need to add.”
• “From Google Search’s perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO.”

Google: Optimizing your website for generative AI features

So the checklist is the old one: readable HTML, clear structure, accurate metadata, a site that loads quickly, and content that answers the question. Anyone selling llms.txt as a ranking lever is selling something Google says it ignores.

Google also documents a Search Console inclusion control and a Generative AI performance report, with worldwide rollout noted as complete on 31 August 2026. Check the actual property’s setting and inheritance. The Search report shows impressions for AI Overviews and AI Mode, not a separate breakdown of AI clicks or conversions; its data overlaps the ordinary Web report. A low-impression property may not display it. Google inclusion control, Google AI performance report

How much traffic is at stake

Less than the headlines suggest, for now. SE Ranking measured ChatGPT’s share of referral traffic across 101,574 sites at 0.32% in May 2026, up from 0.23% a month earlier. Similarweb reported in September that AI referrals “remain a low single-digit percentage of most sites’ total traffic.” Cloudflare’s July 2026 report put Google at roughly 88% of referral traffic. SE Ranking, Similarweb, Cloudflare: Content Independence Day, one year on

The crawling side is a different story. Cloudflare reports that more than half of all internet traffic is now non-human, and that 52% of crawler requests in June 2026 were for AI training, up from 22% in spring 2025. The crawl-to-visit ratios it published for 2025 were extreme: Anthropic’s crawler fetched as many as 500,000 pages for every visitor it sent, OpenAI’s as many as 3,700 to one, Google’s about 30 to one. Secondary reports citing Cloudflare’s mid-2026 data suggest the AI ratios have fallen into the hundreds and low thousands, still far above search. Cloudflare Radar 2025 year in review

The honest summary: AI systems read a great deal and send very little, and Google still decides most of what a business gets. Making a site readable to AI crawlers is cheap insurance and good practice. Reorganising a business around AI referrals is premature.

Controls

Site owners have more say than they did a year ago. Cloudflare changed its defaults on 15 September 2026 so that new domains block AI training and agent crawlers on pages that show ads, while allowing search crawlers. Its Content Signals policy, adopted on more than 3.8 million domains through managed robots.txt files, lets a site state separately whether its content may be used for search, as AI input, or for training. Cloudflare can also serve a Markdown version of a page to agents that ask for one, though that feature requires a Pro plan or above. Cloudflare: AI crawl options, Content Signals Policy, Markdown for Agents

None of these depend on which framework built the site. They depend on where it is hosted and whether anyone has made the decisions.

The ways to connect WordPress and Astro

Every option for putting Astro in front of WordPress comes down to one question: how does Astro read the content? There are four answers in 2026, and the state of each changed this year.

The core REST API

WordPress includes a REST API for reading content such as public posts and pages. Access rules, authentication, custom post types and field registration determine what a particular site exposes. Astro’s WordPress guide demonstrates fetching content through that API. Confirm the required fields, hosting restrictions and draft access on the source site rather than assuming every installation returns everything the new frontend needs. Astro docs: WordPress, WordPress REST API handbook

Its limits are practical rather than fatal. Menus need authentication to read. Block content arrives as rendered HTML rather than structured blocks. There is no built-in preview handoff and no feed of what changed since the last build. Each of those is a small piece of work in a real project, and none of them is packaged for you.

WPGraphQL

WPGraphQL gives WordPress a typed schema that a developer can query precisely. It reports 30,000+ active installs, is maintained by Jason Bahl at Automattic, and describes itself as “becoming a Canonical Plugin on WordPress.org.” Version 2.23.0 shipped on 15 September 2026, with companion plugins for Advanced Custom Fields, SEO data, and caching. WPGraphQL on WordPress.org

It is a sound choice when a team wants a strict schema. For scale, 30,000 installs is about 1.5% of the two million sites running Advanced Custom Fields, so it is a developer’s tool rather than something most sites already have.

Faust.js and the WP Engine toolkit

Faust.js is WP Engine’s headless framework. It is built for Next.js, and its own site still calls it “The Next.js Headless Toolkit for WordPress.” A proposal to add Astro support was opened in July 2025. As of 22 September 2026 no Astro package exists on npm, and the proposal’s last comment is from August 2025. WP Engine’s newer HWP Toolkit includes one Astro example among many Next.js ones. Faust.js multi-framework proposal, HWP Toolkit

10up’s HeadstartWP, the other agency-backed headless framework, remains Next.js only. HeadstartWP

Community Astro packages

This is where the picture changed most in 2026, and not in the direction you might expect.

Sources: wp-astrojs-integration, dewp, WP Astro MCP, hatch, AstroWP, npm downloads API.

These projects give a mixed picture at the September 22 research cutoff. The package written by an Astro core team member was retired that summer, while the reviewed set also included newer AI-assisted tools and small starters. We did not find an official WordPress loader from Astro or Cloudflare in the sources reviewed. EmDash was the WordPress-related project identified on Astro’s roadmap. This sample does not establish the state of every integration or where all development effort is going.

What that means for a rebuild

Reading content out of WordPress is a solved problem. A competent developer can do it in a day with the REST API, and an AI coding agent can do it in an hour. That is why the packages that only do that have not found an audience.

What a real project has to design is everything around the content: how editors preview a draft on the new site, how a published change reaches the public page without a full rebuild, how every old URL is redirected, how titles, structured data, and sitemaps match what the SEO plugin produced, where forms send their submissions, and how menus and reusable blocks are rendered. Astro 7’s route caching made the publish-and-refresh step possible without a rebuild. The rest is project work, and it is the part worth paying for.

Who moved in 2026, and how

We collected the WordPress-to-Astro write-ups published this year by agencies, developers, and hosts, and read them for what people actually did rather than what they recommended.

Sources: Askan, HostWP, Spiteri, Oui Digital, Fernside Studio, Erickson, InstaWP. The agency cost and hosting figures are the authors’ own claims and we have not verified them.

Astro’s own monthly roundups tell the same story. Every WordPress mention on the Astro blog in 2026 is a migration story, from “Zero Maintenance, Maximum Control” in January to “a move done in less than an hour with Claude Code” in June. What’s new in Astro, June 2026

Two things stand out.

Several developer-led projects in this sample left WordPress entirely. Moving content into repository files can suit some teams. The right choice still depends on who edits, reviews and maintains the site, and which features must survive. The reported timelines apply to those authors’ projects; they do not establish how long another migration will take, with or without AI assistance.

Keeping WordPress is a distinct option for teams that depend on its editor. Several reviewed agency and hosting examples use this arrangement. The selected write-ups do not establish its share of all real-world migrations or why people choose to publish about one approach rather than another.

The provider prices and starter-kit adoption figures above are vendor anecdotes, not a representative market sample. They do not establish what a typical migration costs or whether most buyers prefer a service to a product. Build a budget from your actual requirements and quoted responsibilities.

What this means for your website

This is WP Forged’s reading of the evidence above. It is advice, not a rule.

If developers maintain the content, consider whether you still need WordPress. Files in a repository can suit a small site, but check scheduling, editorial review, media and other dependencies first. Familiarity with code does not by itself make a CMS unnecessary.

If a team depends on WordPress editing, evaluate keeping it while changing the public front end. A headless setup can preserve selected content workflows, but preview, scheduling, permissions, caching and delivery need an explicit implementation and testing. Existing URLs can often stay; changed URLs need relevant destinations. Define those decisions, SEO output, form handling and recovery in the written scope. The architecture decision guide compares this with optimizing conventional WordPress and static export. The rebuild service explains the work we scope.

If your site runs a store, memberships, or logged-in areas, plan those case by case. WooCommerce checkout, subscriptions, and member content live in WordPress and there is no reason to move them. A public front end in Astro can coexist with them, but each one is a decision, not a default.

If your site was built with a page builder, expect the layouts to be rebuilt rather than copied. Elementor, Divi, Beaver Builder, and WPBakery store layouts in their own formats and rely on their own scripts on the public page. A rebuild recreates the design in templates; it cannot import the builder’s layout. That is the single most common surprise in these projects, and it is why we wrote a separate guide for each builder.

If your site is small, your editors are happy, and Google’s report shows it passing Core Web Vitals, stay put. Check the Core Web Vitals report in Search Console before anyone tells you the site is slow. About half of WordPress sites pass. If yours does, the case for a rebuild has to be made on something other than speed.

For the next decision, use the cost worksheet and portable project brief to compare the same requirements across providers. The SEO migration inventory helps assign URL decisions and launch checks before anything moves.

What to watch before the end of 2027

• WordPress 7.2 on 9 December 2026, with Matt Mullenweg leading the release. Watch whether real-time collaboration returns and what the AI features become.
• EmDash 1.0. Cloudflare says it is coming without saying when. A 1.0 with a plugin ecosystem would change the calculation for sites that need few plugins. Watch the registry’s catalogue size, not the version number.
• Astro’s CDN cache integrations leaving experimental status. When Cloudflare’s provider is marked stable, the publish-and-refresh loop for WordPress content becomes fully supported rather than mostly supported.
• Whether anyone ships an official WordPress loader for Astro. Faust.js promised one in July 2025. Astro’s core team retired theirs in July 2026. If either Cloudflare or WP Engine ships one, the bridge becomes a commodity and the value moves entirely to the project work around it.
• W3Techs’ monthly figure. WordPress crossed below 41% in September 2026. The next line is 40%.
• The WP Engine trial in October 2027, and any settlement before it. The outcome affects who controls Advanced Custom Fields and how WordPress.org is run.
• New measurements of AI crawler behaviour. The December 2024 study cited here is historical. Prefer current, reproducible observations about a named crawler over a blanket rule about all AI products.

How we researched this

Download counts, plugin install counts, version numbers, and repository activity were pulled from the npm, WordPress.org, and GitHub public APIs on 22 September 2026. Core Web Vitals figures are from HTTP Archive’s technology report for the 1 August 2026 dataset. Market share is from W3Techs on the same day. Every dated claim about a release, announcement, or policy was checked against the primary page linked beside it. Where a figure comes from a vendor’s own marketing, including Askan’s cost claims, Astro’s site traffic, and Limy AI’s crawler statistics that we chose not to use, the text says so. Interpretations are marked as WP Forged’s reading. WP Forged rebuilds WordPress front ends in Astro, and readers should weigh our conclusions with that in mind.

Appendix: the numbers

A. npm downloads, August of each year, and the latest 30 days

Latest 30 days is 23 August to 21 September 2026. Astro’s share of Next.js downloads in August: 2.6% (2023), 4.1% (2024), 4.7% (2025), 8.9% (2026); 9.7% in the latest 30 days.

B. WordPress-adjacent npm packages, latest 30 days

Latest versions: astro 7.3.3 (16 Sep 2026); emdash 0.38.0 (15 Sep 2026); @faustwp/core 3.4.1 (4 May 2026); wp-astrojs-integration 0.5.1 (23 Mar 2026).

C. WordPress.org active installs, 22 September 2026

WordPress.org reports installs as bucketed floors.

WordPress version share, per WordPress.org on 22 September 2026: 7.1 56.3%, 7.0 12.5%, 6.9 7.7%, 6.8 5.8%, 6.7 2.7%. PHP 8.x runs 77.3% of reporting sites; 16.8% still run PHP 7.4.

D. Repositories, 22 September 2026

E. Core Web Vitals by technology, HTTP Archive, dataset 1 August 2026

April 2026 for comparison: all origins 54.2%, WordPress 49.1%, Astro 67.8%.

Figures and feature status were checked against the primary sources linked throughout this report on 22 September 2026. Interpretations are identified as WP Forged’s perspective. Later releases and later months of data will change the numbers described here.