Skip to main content
Grants up to $15,000/year for media
InclusiveWeb

AI Alt Text · Beta

AI writes alt text from the page context — and applies it with no code changes

Images without alt text are the most common accessibility failure on the web. AI writes the description from the page context, and it reaches your site on its own, with no change to your code.

Included in Autofix. Nothing to buy separately.

How it works

Four steps, and none of them needs access to your code.

  1. 1

    Discovery in a real browser

    The script sees the images on a page a real visitor opened. That brings in pages a crawler would never reach — product cards, search results, filtered listings.

  2. 2

    Page context, not just the picture

    Along with the image the model gets the heading, the surrounding text, the caption, the link it sits in and the page language. So the description explains the image’s role here rather than listing what is drawn on it.

  3. 3

    Generation

    A vision model returns one short phrase, up to 100 characters, in your language. One credit is one generation; a regeneration spends a credit too.

  4. 4

    Applied on the site

    The finished description is served to later visitors by the same script that carries the rest of Autofix. Nothing changes in your code.

Honest about the first visitor

An image nobody has opened yet has no description — that visit is what discovers it. The next visitor sees the alt text. For pages in the scanned sample the description is ready in advance.

Why this is not a caption

The model follows an instruction written for WCAG, not for a pretty description. Here is what it says.

  1. 01

    Purpose, not appearance

    The description conveys what is shown and why it is here. Colours, shapes, object counts and background are not.

  2. 02

    No "image of…"

    The screen reader has already announced that this is an image. Repeating it in the alt steals the user’s time.

  3. 03

    An icon is its function

    "Play video", not "blue triangle in a circle". The user has no use for the geometry.

  4. 04

    A person is an action or a role

    What they are doing, not how they look. No gender, age or appearance unless that is the point.

  5. 05

    It does not repeat nearby text

    If a heading or caption sits next to the image, the description does not say it a second time.

  6. 06

    Short

    One phrase, up to 100 characters. A long alt is read out in full and cannot be skipped.

How many generations are included

Credits follow your Autofix tier and last for the whole subscription — they do not expire monthly.

How many generations are included
Site sizeGenerations
up to 50100
up to 500500
up to 5,0001,000
up to 50,0003,000
up to 250,0004,000
Trialevery image on 3 pages

When credits run out, generation stops and the records move to manual entry — writing the description yourself in the dashboard is free and unlimited. As soon as credits refresh, the AI returns to those images.

What it covers, and what it does not

Covered

  • Ordinary `img` elements with no alt or an empty alt
  • Inline `svg` and elements with `role="img"` — via `aria-label`
  • Image-map `area` elements and `input type="image"` buttons
  • `object` elements — via `aria-label`

Not covered

  • CSS background images — they are not content in the DOM
  • `canvas` and graphics drawn by script
  • Complex infographics that need a long description alongside
  • Cases where the right answer is an empty alt because the image is decorative

The criterion is WCAG 1.1.1 Non-text Content, Level A. It is the most common accessibility failure on the web and one of the few where automation can genuinely create the right content rather than guess the author’s intent.

Frequently asked questions

Do I need to change anything in my site’s code?

No. It runs on the same Autofix script you already added in one line. The alt is set in the visitor’s browser; your CMS and templates are untouched.

Does the description appear immediately?

For pages in the scanned sample, yes — it is ready in advance. For the rest, the first visitor opens the page and thereby sends the image into the pipeline, and the next visitor sees the description. Usually minutes.

Can I fix a description I do not like?

Yes, every description can be edited or regenerated in the dashboard. Text you edit by hand is left alone by the AI afterwards. A regeneration spends a credit; editing does not.

Do credits reset every month?

No. It is a total for the whole subscription, not a monthly quota. We deliberately avoid a monthly reset: most images on a site are described once and never change.

What language does the AI write in?

The language of the page, taken from the `lang` attribute and the context. On a multilingual site each language version gets its own description.

Why Beta?

It works and you can use it, but we are still changing the model’s instruction based on results from real sites. Description quality on niche imagery is the main thing we are still improving.

See it on your own images

One line to install. In the trial the AI describes every image on a three-page sample — enough to judge the quality on your own content.

AI alt text for images — Beta | InclusiveWeb