WordPress’s New AI Assistant Lets You Edit Your Sites With Prompts

WordPress is rolling out a built-in AI Assistant that lives inside the editor and the Media Library, with this the place you write and tweak now becomes the place you can simply ask for changes. The feature is pitched as less copy-paste and more direct action, because it can work with what’s already on the page instead of handing you text to manually fit into a layout.
In the editor, the assistant is meant to rewrite or translate sections, adjust styles, and help build pages without you leaving the workflow. The idea is that you describe the outcome and AI applies edits in context. The main limitation is theme support as these editor changes are tied to block themes, and classic themes won’t see the assistant inside the editor. WordPress emphasized that you don’t need fancy prompt phrasing to use it.
The Media Library update is more concrete about what powers it. There’s a Generate Image entry point for creating images and making targeted edits, with controls like aspect ratio and style so you can keep visuals aligned with your site’s look. WordPress says the image workflow uses Google Gemini Nano Banana models, therefore decreasing trips to external tools for basic generation and edits.
A quieter but important piece is how this plugs into collaboration. In block notes, you can type @ai to request things like headline options or a quick fact check, and WordPress says replies can include relevant links and information from outside sources. In the workflow thw feedback, AI suggestions, and the exact block under discussion stays tied together, instead of getting lost in the side chats.
WordPress says Business and Commerce plan sites can opt in at no extra cost, and sites created with its AI Website Builder get AI tools enabled by default under eligible paid plans. This tells that WordPress is trying to make the CMS prompt-able, so “editing the site” starts to look like directing a system.
One detail it doesn’t spell out as clearly is which text model runs the writing and design side, even as it names the image model, which is the kind of implementation detail power users will care about.
Y. Anush Reddy is a contributor to this blog.



