v1.2.0
— The plugin looks at your images — version 1.2.0 LatestRelease notes
Two new features, and one of them needs the free plugin updated to 3.3.0.
Alt text written by AI from the image itself. An Analyse image button appears on any image: the plugin sends the picture to your provider and gets back alt text, a caption and a description. From the settings you can scan the whole library, which runs in small batches with a progress indicator.
Three things it does that are worth knowing before you press the button:
- It never touches alt text you wrote yourself. It only fills what is empty, both on the front end and in the media screen.
- It recognises decorative images — backgrounds, textures, dividers — and leaves their alt text empty, which is the correct form under the accessibility guidelines. Filling it would force screen reader users to listen to noise.
- It does not pay twice for the same image. The result is tied to the file's content: as long as the picture does not change, no new call is made. Replace the file while keeping the same library item and the plugin notices, and analyses it again.
You need an Anthropic, OpenAI or Google Gemini API key: those are the three providers that accept images. If your default provider is Mistral or Kimi the plugin automatically uses another key you have already configured, and if it finds none it tells you which ones would work — it never sends the request to a model that cannot see the image.
llms.txt fills itself in. In the GEO box of posts and pages there is an Include in llms.txt as pillar content checkbox: flagged content becomes an entry in the file AI assistants read to find their way around your site. The description comes from the GEO fields you have already filled in — Direct Definition, or Quick Answer, or the excerpt — with no AI call and therefore no cost.
It comes from a real file: 606 bytes, a title, a summary and zero links. Technically correct, editorially useless. The section existed but had to be filled in by hand, entry by entry, and nobody did.
Costs. Every analysed image is one call on your own API key, exactly like field generation. You can stop the library scan at any time by closing the page: the work already done stays saved.