Your website, modernised on a Drupal CMS you own, ready for LLMs

X to Drupal is a harness built by QED42 that modernises an existing website on any platform by replatforming it onto a real Drupal CMS. The visual result is preserved, your content migrates across, and the backend is modelled the way a senior Drupal architect would model it: typed components, real listings and taxonomies. It is a paid QED42 offering, delivered as a finished Drupal 11 site that you own and can install on any Drupal 11 host.

A modernisation, and the migration is one step of it. Modernising a website in 2026 means four things: AI inside the CMS, digital sovereignty, an open architecture, and readiness for AI agents. X to Drupal replatforms the site to Drupal 11 and Drupal CMS with the design kept and the content moved into a real content model, and that move is what gets the site to all four. The delivered site ships the Drupal AI module suite with an AI assistant button in the editor, an llms.txt map, a markdown twin of every content page, a read-only JSON:API and robots.txt rules for each named AI crawler. It is handed over as a Drupal Recipe and a database dump that is proven to restore, so it installs on any Drupal 11 host with no vendor lock-in.

A platform instead of a page editor. You get content types and fields, revisions, roles and permissions, media management, and an API over all of it, so your team can publish and extend the site without a developer or a deployment. On top of that the delivered site arrives with Drupal's AI layer configured, structured data emitted from the content model, search foundations carried across, and open-source hosting you choose. The reason most teams replatform is not how the site looks, it is what the current platform will not let them do next.

That is the point of the harness. The design comes across as it is, and we verify the result side by side against your live site at 1440, 768 and 375 pixels wide. We publish the comparison rather than a parity score, because a single number flatters the easy pages. The before and after of our own site is on this page.

They migrate into a real Drupal content model. On our own site that meant 569 article bodies moved across with their formatting intact and 2,007 images brought into Drupal's own media handling. Listings become Drupal Views, categories become taxonomies, and every article becomes a structured node your team can edit.

That is treated as delivery work rather than a later project. The delivered site emits a structured JSON-LD entity graph from the content model, so an assistant reading your pages can tell what your organisation is, what each page is about and how they relate, instead of guessing from markup. It also arrives with the four machine-readable surfaces configured: an llms.txt map of the site, a markdown rendering of every page, read-only JSON for every content item, and named crawl rules for the LLMs. Metadata, clean paths, redirects from your old URLs and sitemaps are configured in the same pass. You do not have to take this on trust: qed42.com is the Drupal site we delivered, and it serves qed42.com/llms.txt, a markdown twin of every page, qed42.com/jsonapi and a robots.txt that names the AI crawlers one by one.

By publishing the same content in formats a machine can consume without scraping. Three things do most of the work: structured data in JSON-LD so entities and relationships are explicit, a plain-text map at llms.txt that tells an assistant what the site is and lists its canonical URLs, and a clean markdown or JSON copy of each page so the prose arrives without navigation, scripts and styling wrapped around it. Migrating to Drupal is the moment to put these in, because the content model that generates them is being built anyway. On a closed platform you usually cannot add them at all.

Drupal's AI module suite is installed and wired to your own model provider keys, so the capability belongs to your CMS and your account rather than to a tool we hold. On delivery that covers assistive authoring inside the editorial experience, alternative text generated for images that were missing it, and page titles, meta descriptions and social metadata drafted per content type from the content itself, with an editor approving rather than starting from a blank field. Two things are still in build and we label them that way: deeper AI authoring across an existing library, such as long-form expansion and summaries applied in bulk, and a deeper automated SEO audit.

You get a runbook instead of a training course. A phase of the run generates a guidebook to the site it has just built, written for someone who does not work in Drupal: the content architecture in plain English, every component shown as it looks once it is populated, where each part of a page comes from and which component holds it, and how to manage taxonomies, menus and listings. It is handed over with the site, so the knowledge does not sit with the one or two people who happened to attend a session, and it is still there six months later when someone new joins.

Yes, and in plain language. The delivered site is wired for a Model Context Protocol connection, an open standard that lets an AI assistant work through an application's own interface rather than around it. Someone on the content team asks for a page, a component or a copy change, and it arrives in Drupal as a draft built through that content type's own rules, so the metadata, the path and the structured data are filled the way the content type already defines them. Nothing publishes itself: the draft stays unpublished, under the site's own roles, until an editor reviews and publishes it.

That is what the brand context is for. The run captures the design system it has just reproduced, the colours, the type, the spacing, the components and the rules that hold them together, alongside the voice the existing site is written in, and hands that over with the site. A change requested a year later is drafted against your own system rather than a generic one, and because it arrives as a draft inside Drupal an editor still sees it before anyone else does.

The harness run itself takes days, and it grows with the size of the site: more pages, page templates and forms make a longer run. Most of the calendar time belongs to people: your team's decisions on what to keep and what to retire, the review of the delivered site on your own pages, and the go-live window, with environments, DNS cutover and a rollback plan. The acceleration section on this page shows four typical site shapes, from a small brochure site to a publisher, each with the length of its harness run and the weeks from kickoff to go-live. For your own site, get in touch through the form on this page and we will come back with what replatforming would mean for yours.

Because on a new platform both the design and the content model have to be recreated by hand, page by page and field by field. The old site works as a reference, and everything else is labour. X to Drupal moves that recreation into the harness run, so the design and the content model come across from the site itself, and the first post in our blog series walks through it in detail.

Search standing is treated as part of the migration. Page metadata, clean paths, redirects from the old URLs, sitemaps and JSON-LD structured data arrive configured on the delivered site rather than left as a to-do list, which is where most replatforming projects lose rankings. The machine-readable surfaces an LLM reads arrive with them, so the site is legible to an assistant on day one rather than after a follow-up project. A deeper automated SEO audit is still in build and we say so plainly rather than counting it as shipped.

Those tools help you build a new site on their platform. X to Drupal migrates the site you already have onto Drupal, an open-source CMS you own and can host anywhere, with the design you already invested in preserved. If your website is an asset rather than a draft, keeping it intact is the difference that matters.

You own it. The deliverable is a standard Drupal 11 site: typed paragraph components, Views, taxonomies and configuration, with zero custom modules carrying the site. It installs on a fresh Drupal 11 as a Recipe, and any competent Drupal team can maintain it, including one that is not us.

Animation-heavy sections are the hardest thing to reproduce exactly, and we show them honestly in the side-by-side rather than hiding them. We also never quote a single parity percentage, because fidelity varies by page and viewport. On the AI side, the layer, the structured data and the machine-readable surfaces are all configured on delivery, while deeper AI authoring across a library and a deeper automated SEO audit are in build rather than shipped, and this page names which is which.

Yes, ours. qed42.com ran on Webflow and was migrated to a Drupal 11 site with X to Drupal: 569 article bodies, 2,007 images, real listings with filters, case studies and forms. That Drupal site is what qed42.com serves today, so the production site and the proof are the same address. Betting our own homepage on it was the fastest way to find out whether the claim survives contact with reality.