The WordPress AI plugin is block editor only

Install the WordPress AI plugin on a site where the classic editor plugin is active, open a post, and look around. There is no AI panel, no toolbar button, and no notice explaining the absence. That silence is where the support requests come from, and it is why this week’s AI team meeting put the plugin’s editor scope in writing.

The decision is that the plugin supports the block editor, and the readme now states it plainly. I think it’s the right call. The plugin’s UI needs the block editor. The AI primitives underneath it do not, and knowing the difference keeps you from telling a client something that isn’t true.

The change came out of the 10 June contributor call, summarized on make.wordpress.org.

Why the WordPress AI plugin only needs one editor

First, it formalizes what was already the case. @dkotter confirmed in the meeting that the plugin has never supported the classic editor, so the readme change removes an ambiguity rather than dropping support for anything. The repo readme already carried the same statement, so the gap was in communication. Nothing in the plugin’s behavior changed. An unstated boundary costs maintainer time in the support thread and again in the answer.

Second, maintainer time is the scarce resource here, and the meeting was unusually open about that. The team walked its full repository list and found that the provider plugins (OpenAI, Anthropic, Google), the PHP AI Client, and php-mcp-schema all lacked maintainers. The PHP AI Client is the urgent one because it is on a path into core with five weeks to Beta 1, and the people who normally commit there are pulled across other work. Given that shortage, going deeper on one editor is the responsible choice. Spreading those hours across three editors would accomplish less.

Third, a boundary only hurts if it walls off the primitives, and this one doesn’t. @justlevine’s endorsement in the meeting was conditional in exactly the right way: a canonical plugin defining its limits is fine so long as the underlying primitives stay available for anyone building classic editor or page builder integrations. Watch that condition over the next few releases rather than the readme line.

Where the AI surface lives

It helps to separate the layers, because the meeting notes cover all of them in one breath. The Abilities API is in core, with the foundational primitives delivered across WordPress 6.9 and 7.0, and abilities filtering plus new lifecycle hooks now in trunk. New core abilities are shaped as plugin experiments first, and their schemas land in the plugin before core for future compatibility. The PHP AI Client is heading into core on the 7.1 cycle. The MCP adapter, which exposes abilities over Model Context Protocol so outside agents can call them, has its 1.0 checklist open on GitHub.

None of that is block editor specific. The editor requirement lives in the plugin’s UI layer, in the code that renders panels and toolbar buttons inside Gutenberg. If you build against the client and the abilities registry and surface the results in your own metabox, a classic editor site works fine. If you’ve ever ported an agent workflow into WordPress, you already know the plumbing is the stable part.

If you want to see the boundary on a real install, this logs every admin screen that is not a block editor screen:

<?php
add_action( 'current_screen', function ( $screen ) {
	if ( isset( $screen->is_block_editor ) && ! $screen->is_block_editor ) {
		error_log( sprintf( 'No block editor on %s, the AI plugin UI will not load here.', $screen->id ) );
	}
} );

Run it on staging and you get a concrete list of screens where the plugin’s UI will never appear. That list usually makes for a shorter client conversation than explaining the architecture.

What I’d tell a client on the classic editor

The other side has a point, and it deserves a fair hearing. The Classic Editor plugin still has a large installed base, as the plugin directory listing itself acknowledges, and plenty of WooCommerce-heavy stacks simplify or replace the editor outright. Editorial teams, the same ones still installing classic editor hotfixes for WordPress 7.0, are not going to switch to blocks to get an AI panel. For those sites the official path ends here, and the realistic options are a commercial product or custom work.

There is a decent sign for them in the same meeting, though. Automattic and Elementor contributors talked about building on shared primitives like theme.json, which is the same pattern: get the underlying APIs right and vendors can ship their own integrations on top. If the abilities and client layers hold that promise, builder vendors and agencies can cover what the canonical plugin won’t.

So for a store still on the classic editor, my quote would not include the official plugin. It would include work against the AI client and the abilities registry, with UI placed where those editors already live. And when clients go looking for their own answers, note that user-facing docs are moving off the team handbook onto the WordPress.org documentation site, which should eventually make the question of editor support easier to find a straight answer to.

If you’re holding client sites on the classic editor or a page builder and want AI features built on the core primitives rather than the block editor UI, that’s the kind of work I take on, and I’m glad to scope what it would look like for a specific stack.

Run wp plugin list --status=active --fields=name | grep -E 'classic-editor|elementor|ai' across your sites and count the ones where the official plugin can do nothing.

author avatar
Ahmad Wael
I'm a WordPress and WooCommerce developer with 15+ years of experience building custom e-commerce solutions and plugins. I specialize in PHP development, following WordPress coding standards to deliver clean, maintainable code. Currently, I'm exploring AI and e-commerce by building multi-agent systems and SaaS products that integrate technologies like Google Gemini API with WordPress platforms, approaching every project with a commitment to performance, security, and exceptional user experience.

Leave a Comment