WordPress AI features just got a massive upgrade with the 0.7.0 release of the experimental plugin. I’ve been following the AI-powered integrations in core for a while now, and this update is more than just a minor bump. Specifically, we’re seeing the foundation for how WordPress 7.0 will handle metadata and editorial logic. It’s messy, it’s experimental, but if you’re managing large content libraries, you need to pay attention.
Content Classification and SEO Automation
One of the biggest bottlenecks in any editorial workflow is taxonomy management. I’ve seen sites with 5,000 tags where half are duplicates like “WP” and “WordPress.” Furthermore, manual tagging is the first thing writers skip when they’re under a deadline. The new Content Classification experiment tries to fix this by suggesting Categories and Tags based on your content, title, and excerpt.
In contrast to previous “magic” solutions, this version limits suggestions to preexisting terms on the site. This is a massive win for consistency. If the AI suggests a tag that doesn’t exist, it’s usually just noise. By tethering the model to your actual taxonomy, the dev team is ensuring we don’t end up with a database full of “AI-slop” tags.
Bulk Alt Text: The Accessibility Bottleneck
We need to talk about the Media Library. For some reason, the standard advice has been “just add alt text later,” and consequently, most sites have hundreds of images with zero accessibility metadata. Version 0.7.0 introduces Bulk Alt Text Generation. You can now select multiple images and let the AI handle the descriptions based on the W3C Alt Text decision tree.
However, don’t just “set it and forget it.” AI descriptions can be hit or miss depending on the provider you’re using. If you’re on a legacy host with limited resources, running these bulk jobs might hit a transient timeout. I always recommend checking the background processing logs if you’re doing more than 50 images at a time.
Extensibility and the 0.7.0 Developer API
As a developer, the most interesting part of this release isn’t the UI—it’s the hooks. The team has introduced ways to modify system instructions and adjust results before they hit the database. If you’re building a custom editorial dashboard, these hooks allow you to inject brand voice guidelines directly into the AI’s context.
Earlier this year, I discussed future-proofing WordPress AI features, and 0.7.0 is exactly what I was hoping for. It moves away from “black box” generation and toward a developer-friendly infrastructure.
<?php
/**
* Example: Customizing the AI Post Context
* Prefix: bbioon_
*/
add_filter( 'wp_ai_post_context', function( $context, $post_id ) {
// Inject specific brand tone guidelines into the prompt context
$brand_guidelines = get_option( 'bbioon_brand_voice_rules' );
if ( ! empty( $brand_guidelines ) ) {
$context['system_instructions'] .= "\nFollow these voice rules: " . $brand_guidelines;
}
return $context;
}, 10, 2 );
Look, if this WordPress AI features stuff is eating up your dev hours, let me handle it. I’ve been wrestling with WordPress since the 4.x days, and I can help you build a workflow that actually scales without breaking your site.
The Road to 0.8.0 and Beyond
The roadmap for 0.8.0 already looks ambitious, including “Refine from Notes” and Dashboard widgets. The key takeaway here is that AI is no longer a plugin you just “try”; it’s becoming a core architectural layer. If you haven’t tested the v0.7.0 release on a staging site yet, now is the time. Just make sure your PHP environment is at least 8.1 to avoid race conditions with the newer async providers.
