Drupal AI in 2026: Status, Architecture, and Roadmap

On April 24, 2026, I gave a session at Drupal Developer Days Athens titled Status, Contribution, and Roadmap for Drupal AI in 2026. This post is the written version for people who weren't in the room and for the organizations weighing how to get involved.

One year ago, Drupal AI was a strategy conversation. It isn't anymore.

According to Drupal.org usage statistics, the AI module reached 13,980 active installs by April 12, 2026. The curve is steep and adoption is still climbing.

How the Drupal AI Initiative Is Structured and Funded

The Initiative formed in June 2025 with a Leadership Team of five companies. Work is organized into two parallel workstreams. The Innovation workstream handles R&D, prototyping new capabilities. The Product workstream stabilizes mature features and ships them to production. Product Managers are appointed by Leadership, and each workstream has a Product Owner who runs the two-week sprints.

The four people to know if you want to contribute are Rakhi Mandhania (Innovation Product Owner), Arian Raeesi (Product Product Owner), and the technical leads Marcus Johansson and Artem Dmitriiev. I serve as Product Manager for the Drupal AI Product workstream, sponsored by 1xINTERNET, one of the co-founding companies of the Initiative.

More than 30 Partner organizations contribute to the Initiative today, with over 25 full-time-equivalent staff committed across development, professional UX, professional QA, and marketing. 

The reason for the high velocity is that the Partners sponsor the work, with money and with their people. If your organization wants to be part of it and shape what gets built, become a Partner.

Drupal AI 2.0 Architecture

The architecture of Drupal AI 2.0, the next major release, has a clear separation of concerns.

The main components are AI Core, the AI Official Modules, and the AI Official Recipes. Drupal AI depends a number of important general purpose modules and has a large contrib modules space.

At the bottom is Symfony AI, the vendor abstraction layer that lets Drupal AI talk to OpenAI, Anthropic, Gemini, and others without leaking provider-specific code into the rest of the stack. AI Core sits on top and is vendor-agnostic. It contains the foundational pieces every AI feature relies on: Agents, Automators, Guardrails, Moderation, Observability, and the APIs that tie them together.

User-facing features live in the Official Extensions: Initiative-supported modules like AI CKEditor, AI Search, and AI Chatbot. Recipes provide ready-to-use configurations of this functionality, such as Prompt Safety and Image Classification.

Naturally, AI functionality relies on the wider Drupal ecosystem. The Ecosystem Dependencies are general-purpose modules the AI stack builds on, including Context Control Center, Field Widget Actions, MCP, and Tools.

Beyond the Initiative-maintained layer, Contrib Modules add community-maintained integrations and tools.

Drupal AI: Three layers, one platform.

Recipes: The Fastest Path From Idea to Running Site

A Drupal Recipe bundles configuration and demo content into something a site builder can apply with a single command. For Drupal AI, Recipes are both the delivery mechanism and a way to learn how the functionality is put together. You can apply a recipe as is, study its configuration to see how Drupal AI is composed, or adapt it for your own project. Going from "I want to try this" to "it is running on my site" takes one command, and what you get afterwards is a working reference you can review.

ddev drush recipe ../recipes/<name>

Below are two examples from the session.

AI Image Classification configures AI Automators together with Field Widget Actions. Upload an image, and alt text, an image description, and tags are generated automatically. The recipe is a concrete reference for how these two modules are used together.

ddev composer require drupal/ai_recipe_image_classification

ddev drush recipe ../recipes/ai_recipe_image_classification

AI Guardrails ships as two reference configurations. PII Guardrails intercepts requests and responses to block leaks of email addresses, phone numbers, IBANs, and credit card numbers. Prompt Safety Guardrails covers liability and security concerns including legal and medical advice, dangerous HTML, and code injection patterns.

ddev composer require drupal/ai_recipe_guardrails_pii

ddev composer require drupal/ai_recipe_guardrails_prompt_safety

ddev drush recipe ../recipes/ai_recipe_guardrails_pii

ddev drush recipe ../recipes/ai_recipe_guardrails_prompt_safety

Built With Drupal CMS, Not Around It

Drupal AI is not developed independently. Drupal itself is a platform, and Drupal CMS is a product built on top of it that packages the most common functionality so site builders get a working starting point. Drupal AI is developed alongside Drupal CMS: our demos run on Drupal CMS, and roadmap and UX are coordinated with the Drupal CMS team. The AI Dashboard ships since Drupal CMS 2.0, which means every Drupal CMS user has a single place to discover and apply the recipes the Initiative provides. The two products move together.

The 2026 Roadmap

The Drupal AI 2026 Roadmap is a community-aligned plan with eight points. It was introduced by Dries Buytaert, with the full roadmap document available alongside the blog post.

This is the contract. Everything the Initiative prioritizes for the next 12 months has to map back to one or more of these eight points.

What We Are Building in the Next 3 to 6 Months

The Innovation workstream is focused on three things. MCP Client and Server turns Drupal into a tool that external AI hosts like Claude, Cursor, n8n, and Activepieces can read from and write to. Canvas AI generates brand-compliant content from your design system components. Migrate to Symfony AI modernizes the framework layer and aligns Drupal AI with the wider PHP and Symfony ecosystem so we maintain less and inherit more. As Innovation features mature, they move into the Product workstream.

The Product workstream is focused on stabilizing and shipping. AI Content Reviews is the closest to release and ships in the coming weeks. It brings scalable AI-assisted governance to editors with flagged issues in plain language and concrete suggestions for improvement.

The first iteration enables four criteria. Brand compliance checks voice, tone, and style. Brand name and term spelling catches things like "DrupalAssociation" when it should be "Drupal Association". Legal compliance flags prohibited claims and missing disclosures. Reading level assesses content for clarity. Each review produces a score per criterion, an explanation, and concrete suggestions for improvement. Future iterations will let editors apply suggestions with one click and will plug into the Context Control Center to source brand guidelines, legal requirements, and editorial standards.

AI Search moves Drupal from listing results to answering questions with synthesized summaries grounded in site content. AI Observability delivers a full audit trail on every prompt and response, with token usage, cost tracking, and OpenTelemetry export.

Two pieces of cross-cutting work sit alongside both streams. The AI Demo System gives Partners flexible demos they can adapt for client conversations. A growing training and webinar program, together with AI coding best practices for agentic Drupal development, equips Drupal developers to build AI applications with confidence.

Each of these maps back to specific points on the eight-item roadmap. MCP underpins Advanced Governance, Intelligent Site Improvements, and Multichannel Campaigns. Canvas AI delivers Page Generation, Context Management, Design System Integration, and parts of Content Creation and Discovery. AI Content Reviews ties into Context Management, Background Agents, Advanced Governance, and Intelligent Site Improvements. AI Search delivers the rest of Content Creation and Discovery and feeds into Multichannel Campaigns.

How To Participate

There are two ways in. The first is partnership. Drupal AI Partners pledge people and funding, set priorities at the regular partner meetings, and shape the backlog. If that is your organization, become a Partner.

The second is direct contribution. The prioritized backlogs are available on Drupal's GitLab instance (requires Drupal user account), and any Drupal contributor can pick up issues without being a Partner. The people to talk to about contributing are Rakhi, Arian, Marcus, and Artem, all linked above.

For the longer story of how the Initiative came together, From Leuven to Athens is the one-year retrospective.

A year ago this was an idea and a rough roadmap. Today it is a product I get to help ship every two weeks, and I cannot wait to see where the next year takes it.