FeaturesPlans
ITENBuy now

AI alt-text with Vision

What it is

AISA writes the alt-text for your images with AI. With Vision turned on, the AI actually looks at the image and describes what is shown; with Vision off, it infers the alt-text only from context (filename, title, caption, parent page). Vision needs a vision-capable provider: Claude, GPT (OpenAI) or Gemini — DeepSeek and Mistral have no Vision model and always fall back to context. Alt-text is capped around 125 characters, descriptive, with no "image of" filler, and written in the language detected from the caption/title (or the site language).

Where to find it

Turn Vision on in AISA → Settings → API Keys → Images & Vision ("Vision active — the AI analyzes the image"). Generate alt-text from AISA → 1-Click SEO (tick "Images (alt-text)") or from the Images tab in Bulk SEO.

For one image while you are writing there is Describe with AISA in the block editor: select an image block and the button is in the block toolbar and in the sidebar, right next to the Alternative text field. Same generation, same cost, same result as the Media Library button — it has its own article.

You can also stop asking altogether. In the AISA Vision section of the same tab, Describe new images on upload makes it automatic: every image you add — file picker, drag and drop, block editor — gets its alt text a few seconds later, in the language of your site, without you opening it. It is off by default, so nothing is spent until you switch it on, and it has its own article.

Requirements & tier

Requires at least one vision-capable provider configured (Claude / GPT / Gemini) in BYO-key mode, or Cloud credits. DeepSeek/Mistral only run the context fallback. Estimated Vision cost is roughly €0.0001–0.004 per image.

How to use

  1. Open Settings → API Keys → Images & Vision and pick the dedicated image provider.
  2. Make sure "Vision active" is checked.
  3. Open the Images tab in Bulk SEO, or run 1-Click SEO with "Images (alt-text)" ticked.
  4. Review the estimate, then start; the AI sees each image and returns alt + title.
  5. The alt-text is saved on the attachment and, in the content, on the matching <img> tags. In the Images tab of Bulk SEO the Alt text column shows the text that has just been written in full, not just a green check, so you can read the results row by row while the run goes on.

Questions people ask

  • Does DeepSeek support Vision? No — neither does Mistral. They use the context-only fallback.
  • What language is the alt-text written in? The one detected from the title/caption, falling back to the page language and then the site locale.
  • How long is the alt-text? Up to about 125 characters, descriptive, no "image of".
  • What if Vision fails on an image? AISA automatically falls back to the context-based prompt for that image.
  • Which images are sent? A resized version (large/medium) under 5 MB, to keep payload and cost down.

Troubleshooting

If alt-text comes out generic, your provider may not be vision-capable — check the Images & Vision box and switch to Claude/GPT/Gemini. If nothing generates, confirm the provider key passes "Test connection" and that the image file is readable and under 5 MB.

Related

Describe with AISA in the block editor (Gutenberg) · AI image titles · Alt-text for images inside the content (builders) · Product Image SEO Template.

← Back to support