A proposal has been published to merge a new wp_knowledge custom post type into WordPress 7.1, with Guidelines serving as the first feature built on it. If adopted, the proposal would introduce a new primitive for storing author-facing and agent-facing site knowledge in WordPress, including support for knowledge types such as guidelines, memories, and notes.
Guidelines move toward inclusion in WordPress Core
The Guidelines feature was proposed by the AI Team earlier this year and introduced as an experimental feature in Gutenberg 22.7. At the time, James LePage described Guidelines as, “ A single place in WordPress to capture site-wide content standards and context, so publishing tools can manage content that stays on-brand and consistent.” With the new proposal, contributors are seeking to merge the feature into WordPress Core as part of the WordPress 7.1 release.
Explaining the rationale behind the proposal, Greg Ziółkowski wrote that while most sites already maintain content standards, they are often managed outside WordPress. He said writers, editors, plugins, and AI assistants all rely on “persistent and structured knowledge about the site,” but that “today there is no shared place to keep it.” To address this, the proposal calls for a “canonical storage primitive for author-facing and agent-facing site knowledge” that would serve as the foundation for Guidelines and other knowledge types.
Ziółkowski clarified that the proposed merge is limited to knowledge storage and access features. The proposal excludes AI providers, models, retrieval algorithms, and autonomous memory systems, while leaving several related areas for future discussion, including retrieval mechanisms, session state, cross-site user preferences, and encryption at rest.
The proposal defines a new structured knowledge system for WordPress
If approved, the proposal would introduce the wp_knowledge custom post type, the wp_knowledge_type taxonomy, REST API endpoints for working with knowledge records, and a Guidelines settings screen for managing guideline records and scopes. The proposed system would include three built-in knowledge types: Guidelines for storing site standards and editorial rules, Memories for preserving approved contextual information, and Notes for private working content. Plugins would also be able to register additional knowledge types through the proposed registration system.
The proposal also outlines a permissions model for knowledge records. Rather than being treated as publicly accessible content, the records would be managed through authenticated WordPress interfaces and APIs, with new records created as private by default. Under the proposed access model, contributors, authors, and editors would be able to create and manage their own private records, while administrators would have broader access to manage site-wide guidelines and other private records.
Clarifying the scope, Ziółkowski said the proposed infrastructure is not intended solely for AI-related use cases, noting that it is already being used for “plain note-taking and draft syncing” and that the Guidelines feature can benefit any multi-author site seeking consistent standards.
He also emphasized that the proposal does not introduce a WordPress “memory system,” describing memory records instead as user-saved context with a defined access policy. Features associated with memory architectures, including relevance ranking, decay, and consolidation, would remain outside the scope of Core. Ziółkowski added that existing plugins storing AI-related context could continue using their own storage systems or adopt the shared infrastructure.
The core patch is now open for review, and progress can be tracked on Trac ticket #65476. With the broader direction already established, discussion is now focused on whether the proposed API is mature enough for inclusion in Core ahead of WordPress 7.1 Beta 1 on July 15, when its key components would effectively become part of WordPress’s long-term supported architecture.
Feedback can be provided either as a comment on the proposal or on the #core-ai Slack channel.