Alt text has a reputation as an accessibility obligation — something you add because you should. That framing is why most of it is bad, and why a lot of it is missing.
Who is actually reading it
Someone using a screen reader, first and foremost, and they deserve a description that tells them what is in the picture rather than the word “image”. But the platforms read it too, and they use it to work out what a post is about — the same job the caption and the tags are doing.
So the accessible answer and the practical answer are the same answer, which is rare enough to be worth noticing.
What a good description does
- Says what is there, plainly: “A dandelion seed head against a dark sky”
- Includes the text that appears in the image, because otherwise it is invisible to everyone who cannot see it
- Stops. Twenty words is usually enough and forty is usually too many
- Skips “image of” and “photo of” — the reader already announced that
Where it goes wrong
Stuffing it with keywords. It is the oldest trick and the easiest to see through: a description that reads like a search query serves nobody, helps nothing, and makes the post worse for the person it was meant for.
Instagram generates alt text automatically when you leave it blank, and the result is generic — “may be an image of one person and text”. That is what a system guesses when nobody told it. Twenty words from you replaces it.
Video has the same problem, differently
The equivalent for video is captions, and the case is even stronger: most video is watched without sound. Burned-in or platform-generated, either way, a video without them is one that a large share of people scroll past without knowing what it said.