Drupal's headless Canvas, AI tooling and advocacy push aim to close its reputation gap

At DrupalCon Rotterdam 2026, Drupal founder Dries Buytaert used his annual State of Drupal keynote to press a single argument: the platform is now, in his words, "light-years ahead of its reputation," and closing that gap is the project's top challenge. In a blog post summarizing the talk, Buytaert framed the keynote around two years of work that began with Drupal Starshot and has since reshaped both the CMS product and the underlying framework.

Starshot grew into Drupal CMS

Buytaert said he launched Drupal Starshot just over two years earlier because he was no longer certain the project could innovate the way it once did. The outcome, according to him, answered that doubt. Starshot evolved into Drupal CMS, which combines Site Templates, Recipes, Drupal AI and the Drupal Canvas page builder to make building websites easier.

In parallel, Drupal Core has become a faster and better framework for developers. Buytaert described a reinforcing loop: a stronger core supports a better CMS, while building the CMS has in turn driven improvements to core.

Canvas goes multilingual and headless

The first half of the keynote covered multilingual sites, JavaScript front-ends, headless Canvas and Drupal AI. Drupal Canvas, the project's page builder, previously lacked multilingual support; Buytaert said that has been added, and multilingual sites in general are now easier to set up.

Notably for front-end developers, Canvas code components are written in React, and the developer experience was deliberately designed around the tools, patterns and workflows that JavaScript developers already know, including modern AI-assisted workflows, without requiring them to learn Drupal-specific concepts.

Buytaert also introduced Drupal Canvas Headless at the conference. Teams can manage content in Drupal while building the front end separately with frameworks such as Next.js, Astro, TanStack, Angular or Nuxt. Editors keep Canvas's visual authoring experience and can preview exactly how pages will appear on the external front end. According to Buytaert, this removes a familiar trade-off: teams no longer choose between a modern JavaScript front end and visual page authoring.

AI as an interface to Drupal

Drupal AI is moving so quickly that Buytaert said it could have filled the entire keynote on its own. A new preconfigured demo was launched so people can try it directly. Current capabilities include answering questions grounded in a site's own content, auditing content against brand guidelines, translation, classification, and building pages and React components with AI. Through MCP, external AI assistants can work with Drupal content from outside Drupal.

Underneath these features, Buytaert positioned Drupal AI as an "AI harness": organizations connect AI models to structured content, tools and workflows while keeping control through permissions, guardrails, observability, metering and a choice of models.

One demo showed an editor asking an AI assistant to unpublish a page without knowing where to click or understanding revisions, moderation workflows or permissions. The assistant translated intent into action, while Drupal still checked whether the editor was allowed to unpublish and applied the site's publishing rules as it would for a manual action. Scott Falconer, who worked on the demo, recorded a behind-the-scenes walkthrough.

The demo rests on the Tool module, which Buytaert called one of the most important modules for Drupal developers to watch. It provides a standard way to expose existing module capabilities, such as publishing, user management, media processing and configuration changes, so AI assistants and other software can discover and use them. Exposing capabilities still requires too much code today: Buytaert said adding tool support to his own album module took roughly 1,000 lines, even though it has already changed how he manages the more than 10,000 photos on his site.

To reduce that burden, Buytaert is working with Matt Glaman on a Tool API change that lets developers expose existing PHP methods as tools using attributes. In his album module, about 1,000 lines of integration code shrank to roughly 20 attributes. The same tools can also be called by other applications over HTTP, used to generate schemas for JavaScript components, or connected to workflow systems such as ECA, FlowDrop and Maestro.

Paying credit for advocacy

The keynote closed on the reputation gap itself. Progress does not automatically change what people, or AI agents, think of Drupal; Buytaert noted that a coding agent he tested in June did not mention Drupal CMS or Site Templates when evaluating the platform. He announced the Drupal Advocacy Program, led by the Drupal Association, under which advocacy work can earn contribution credit, with additional weight for work that reaches people outside the existing community. Contribution credits already influence visibility in the Drupal marketplace, but the system has historically rewarded code and funding far more than evangelism.

Why it matters

For developers and teams choosing a CMS, the keynote is a substantial data point. The announcements target long-standing objections to Drupal: headless Canvas supports modern JavaScript architectures without giving up visual authoring, React-based components lower the barrier for the world's largest front-end community, and the Tool module plus MCP positions Drupal content and workflows for an AI-assisted future in which assistants execute while the CMS remains the governance layer. The advocacy program is equally telling: Drupal's own leadership believes the platform's reputation lags its capabilities, and that perception, not technology, is now the binding constraint on adoption. Whether the gap closes depends on developers outside the community evaluating the current product rather than their memory of it.