The ID Toolbox
← Guides

Alt text for e-learning images

Which images need a description, which should be silent, and what to write for the ones in between. Plus the nine faults that make alt text worthless even when every image has some, and how to find them across a whole course.

Updated September 15, 2026

Alt text is the sentence a screen reader says instead of showing the picture. Most courses have some on most images, which sounds like the job is done and usually is not: the descriptions are filenames, or they all say the same thing, or the decorative background gets a paragraph while the diagram that carries the point says “image”.

The work is mostly deciding which images need words at all.

Sort every image into one of four piles

1Decorative: say nothing, deliberately

Background photographs, dividers, the abstract swoosh behind the title. If removing the image would cost the learner nothing, it should be marked decorative so the screen reader skips it entirely. Silence is the correct description, and marking it is what makes the silence intentional rather than a gap.

2Informative: one clear sentence

A photograph making a point, an icon that means something, a person demonstrating a technique. Write what the image tells the learner, in the context of this slide, not what it contains.

Good: An operator checking the forks of a forklift before a shift.

Not useful: hero-photo-final-v3.jpg. Or: Image. Or: Photo of a man.

3Complex: describe the point, put the detail in the page

Charts, process diagrams, org structures. Alt text is heard once, cannot be paused and cannot be re-read, so a paragraph in it is a poor experience. Put a short description in the alt text and the actual content on the page or in the notes, where every learner gets it.

Alt: Chart of heat illness cases by month, described in the text below. Then, on the slide: Cases peak in July and August, with 60 percent of the year’s total in those two months.

4Text in an image: avoid it, and if you cannot, repeat it exactly

A slide whose words live inside a PNG is invisible to a screen reader, unsearchable, unselectable and untranslatable, and it becomes a rebuild in every other language. Where it is unavoidable, the alt text has to carry the words verbatim.

Nine faults worth hunting for

These are what actually turns up in a finished course, and they are invisible when you are looking at the slide rather than the description:

  • no description at all, on an image that clearly needs one
  • a description that is only spaces, which reads as described but says nothing
  • the filename, which is the single most common fault
  • a redundant opening: “Image of”, “Picture of”, “Graphic showing”. The screen reader has already said it is an image
  • placeholder text nobody replaced
  • a description long enough to be a paragraph
  • a description too short to mean anything: “photo”, “icon”
  • the same description on several different images, which usually means somebody pasted it down a list
  • an image marked decorative that carries a description anyway, so two people disagreed about what it is for

Writing them well

  • Write for this slide. The same photograph might be “a warehouse in summer” on one slide and “a warehouse with no shade over the loading bay” on the slide about shade. Context decides the sentence.
  • One sentence, usually under 150 characters. If it will not fit, that is a signal the detail belongs on the page.
  • Do not repeat the caption. If the text next to the image already says it, the image can be decorative. Hearing the same sentence twice is worse than hearing it once.
  • Mind the buttons. An icon-only button needs a description of what it does, not what it depicts: “Play the video”, not “triangle”.
  • Check the reading order too. Perfect alt text read in the wrong order is still a bad slide, and reading order is set separately in most authoring tools.

Then listen to it

Turn on the screen reader you have, tab through one slide, and listen. Five minutes of that teaches more than any checklist, including this one: you will hear the filename, the doubled caption and the paragraph that will not stop, and you will never write them again.

Two caveats worth being straight about. A tool that reads a file can tell you that a description is missing, or is a filename, or is duplicated. It cannot tell you whether a description is any good, and nothing can: that is a judgment about meaning. And none of this makes a course conformant on its own. Alt text is one requirement among many, and the rest are not machine-checkable either.

Questions people ask

Does every image need alt text?

No. Every image needs a decision. Informative images need a description; decorative images need to be marked as decorative so they are skipped. What fails an audit is the image nobody decided about.

How long should alt text be?

Long enough to carry the point, which is usually a sentence. Past about 150 characters it stops working as a label; past 250 it is a paragraph a listener cannot re-read. Move the detail into the page text instead.

Can AI write my alt text?

It can draft it, and for straightforward photographs it does a decent job. It cannot know why the image is on that slide, which is the part that makes a description useful. Treat generated alt text as a first pass to edit, not a box ticked, and never ship it unread.

What about video and audio?

Different requirement, same principle. Video needs captions and, where the visuals carry meaning the audio does not, an audio description. Narration needs captions, and reading speed matters as much as their existence.

Read next

Tools in this guide