Social Security Administration Guide: Alternate Text for Images

Redundant alt-text (now hear this… twice)

“Pastel lays down with her harness on. There is a long handle running the length of Pastel’s back. This is the handle that the handler uses to follow Pastel’s movements. The long handle is attached with leather straps that go in front of the chest and behind the front legs.”
It’s redundant. In this case, all of the pertinent information about the harness is contained in the surrounding text. Repeating this text in the picture is (a) cumbersome to read; and (b) forces the reader to have to analyze and compare both the picture text and the surrounding text to know whether they have missed any pertinent information.

“Pastel lays down with her harness on.”

“Pie chart for Pastel’s long-range plan for dog treat distribution at the Agency. Pastel = 99%; Other service dogs = 1%”
Half of the text is already in the caption.

“Pie chart. Pastel = 99%; Other service dogs = 1%”

If the caption is in text form (programmatically attached or adjacent to the image), there is no need to repeat the caption in the alt-text because a screen reader will speak it. If, however, the caption is rendered as part of the image, the alt-text must include the caption.