WordPress released AI Experiments 0.4.0, and the interesting part is not the new buttons. This plugin is where the canonical foundation for AI inside WordPress Core is being worked out, with the WordPress 7.0 roadmap as the target. Media handling and editorial review are the two areas that change most.
Why AI Experiments 0.4.0 matters for developers
I have installed enough “AI for WP” plugins to know the usual pattern: non-standard transients piling up in the database and UI bolted on wherever it fit. AI Experiments 0.4.0 works differently. It sits on the new WP AI Client architecture, which is being refined so it can go straight into the core API, and it ships a Generate Image workflow that runs in both the Block Editor and the Media Library.
Generating an image used to mean leaving the editor or fighting a clumsy iframe modal. Now it happens in place. You write a prompt, tweak it, and insert the result into the block. The images also land in the Media Library as normal attachments, so they follow the standard attachment protocols instead of turning into the orphaned assets that third-party AI tools tend to leave behind.
Generate Review Notes: AI-assisted editorial
The other new experiment in this release is Generate Review Notes. It reads content and flags accessibility problems such as missing alt text, along with readability and grammar issues. If you run high-volume content sites for clients, that is the sort of check that normally costs someone an afternoon of manual dev hours. It also lines up with the Content Guidelines experiment.
Under the hood: the WP AI Client
Read through the source of AI Experiments 0.4.0 and most of the work is on internal APIs. The point is to make AI providers swappable. OpenAI, Anthropic, or a local model wired up through MCP (Model Context Protocol), the UI should look the same either way. That abstraction is what keeps things stable as providers come and go.
The small helper below checks whether a specific AI experiment is active before your own logic runs. It follows the internal naming convention used in the 0.4.0 release.
<?php
/**
* Check if a specific AI Experiment is currently active.
*
* @param string $experiment_id The ID of the experiment (e.g., 'image-generation').
* @return bool
*/
function bbioon_is_ai_experiment_active( $experiment_id ) {
if ( ! class_exists( 'WP_AI_Experiments' ) ) {
return false;
}
$experiments = get_option( 'wp_ai_experiments_settings', [] );
// Check if the specific experiment is enabled in the settings array
return ! empty( $experiments[ $experiment_id ] );
}
// Usage Example
if ( bbioon_is_ai_experiment_active( 'generate-review-notes' ) ) {
// Inject custom editorial logic here
}Roadmap to 0.5.0 and WordPress 7.0
0.5.0 is the release where the plugin drops the experimental label and gets renamed from “AI Experiments” to WordPress AI. The rename is the signal that these features are considered feature-ready. Our notes on WordPress 7.0 AI Updates cover the rest, including the “Try AI” callout that will likely appear on new installations.
Experiments already in the works:
- Contextual tagging, which suggests taxonomies based on post content.
- Image editing in the Media Library, including background removal and canvas expansion.
- Refine from Notes, which actually applies the editorial feedback to the content.
To follow the progress or contribute, there is the official GitHub repository, and we are tracking Beta 2 progress as well.
If AI Experiments 0.4.0 is eating your dev hours, I can take it off your plate. I have been wrestling with WordPress since the 4.x days.
Where this leaves you
AI Experiments 0.4.0 moves AI work out of the plugin-for-everything pattern and into a core-integrated client. That is the part worth paying attention to. Test these APIs while they are still moving, because by WordPress 7.0 they will be the standard way to do this, and learning them on a client site is the expensive option.
