AI wrote your image alt text. Does it explain the purpose?
Check the image in its page context. A description of what is visible can still miss the information or action a reader needs.
Use an AI suggestion as a draft and review it alongside the image, nearby text and intended task. The useful question is what a reader would lose if the image were unavailable.
Describe the job the image does
W3C’s decision tree distinguishes informative, functional and decorative images. A meaningful photo needs a brief description; an image that acts as a link or button needs its destination or action. A purely decorative image can use an empty alt attribute.
The same picture can need different wording
Consider an invented example: a red backpack photographed on a desk. On a product page, “Red backpack with a front zip pocket” may convey a relevant detail. On a page where that photo only decorates a heading and adds no information, repeating the objects may add noise. Decide from the actual page.
Do not squeeze a whole chart into one label
For a complex chart, W3C recommends a short alternative plus a fuller text equivalent. Provide the relevant values or a table nearby, together with the meaning the chart communicates. “A blue bar chart” tells the reader little about its data.
Review the draft against the source
Our review checks three things: any invented person or number, any important omitted label, and any duplicated nearby text. For a button, read the proposed wording without the picture and see whether the action is clear. For a chart, compare the description with the underlying data.
Published October 13, 2026. The backpack is an invented editorial example. We did not benchmark an image model or perform a screen-reader usability study.
Sources: W3C WAI · alt decision tree · W3C WAI · complex images.