WordPress 7.0 is close, and the WordPress AI plugin 0.8.0 landed with a few structural changes worth tracking. Image generation graduated to stable status, and there is now a dashboard widget framework that suggests how site-wide AI insights will eventually reach core.
If you read my notes on AI plugin 0.7.0, the theme there was automation. In 0.8.0 that automation gets more agentic through the new Refine from Notes experiment. Instead of you fixing content by hand from a list of AI suggestions, the plugin applies the editorial notes itself. On a team where consistency is the slow part, that saves real time.
Image generation moves from experiment to feature
The clearest sign of maturity in the WordPress AI plugin 0.8.0 is that Image Generation and Editing moved from “Experiment” to “Feature.” For a canonical plugin, that means the API is stable and a breaking change in the next minor version is much less likely. The legacy DALL·E models are gone too, replaced by newer ones like gpt-image-2.
The dashboard widget framework
The admin screens change as well. 0.8.0 ships the first set of AI dashboard widgets, and the framework behind them is extensible. You can register your own widgets to surface specific AI capabilities or model connector data right on the WordPress dashboard.
The internals got some cleanup too. The bootstrap process now uses str_contains() and str_starts_with() instead of the older string juggling. The part that will bite people is the new wp_supports_ai() check: wrap your custom experiments and features in it so you are not initializing AI code in an environment that cannot run it.
<?php
/**
* Example of checking for AI support before initializing a custom experiment.
*/
function bbioon_init_custom_ai_feature() {
// Check if the environment and plugin version support AI features
if ( function_exists( 'wp_supports_ai' ) && wp_supports_ai() ) {
// Safe to register your custom logic or dashboard widgets
bbioon_register_custom_dashboard_widget();
}
}
add_action( 'init', 'bbioon_init_custom_ai_feature' );
Gutenberg Guidelines integration
The WordPress AI plugin 0.8.0 now respects the Gutenberg Guidelines experiment, which matters for agency work. Instead of generic AI tone, the abilities read your site-wide editorial standards. If your guidelines say “formal tone, no Oxford commas,” the suggestions will mostly stick to it. That is the difference between a tool that writes and a tool that writes the way your brand does.
Performance and refactors
Token usage got attention this round. On larger posts, the “Review Notes” feature used to get expensive because it shipped too much context back to the model. 0.8.0 sends reduced context during processing, so model costs should come down and long-form content should feel faster.
If you have been working with the MCP integration or custom connectors, watch the ai-wp-admin page. The old ai settings page redirects there now, and the new one is built with wp-build, which is a more reliable bootstrap for the plugin’s internal assets.
If the WordPress AI plugin 0.8.0 work is eating your dev hours, I can take it on. I have been working with WordPress since the 4.x days.
Coming in 0.9.0
The 0.9.0 roadmap includes a “Try AI” callout for WordPress 7.0 and content provenance tracking through C2PA. If you have not tried the official GitHub releases yet, spin up a staging site and see what the automated editorial workflows do to your turnaround time.
