Exceptions to 1.1.1: When the Alt Text Isn't the Answer

Last week I was at the National Association of Government Web Professionals (NAGW) to present on my atomic hub-and-spoke model (more on this soon). While I was there, I was talking with Rob Carr from the WebAIM team, and we were doing what two accessibility people do when they get together: complaining. This time it was about educational tools, specifically the people who build them misunderstanding how the exceptions to 1.1.1 work.

Everyone knows 1.1.1 Non-text Content. It’s the first criterion, the most-tested, the one that shows up in every automated scanner output. All non-text content needs a text alternative that serves an equivalent purpose.

The people building educational assessments have a genuine dilemma. If a math quiz asks a student to identify which grouping shows five objects, and the image alt text cheerfully announces “five red balls,” the assessment is broken. Not inaccessible — broken. The text alternative has invalidated the very thing being tested. A student who can’t see the image now has the answer handed to them, while their sighted peers have to count.

The exception permits withholding the answer. It does not permit withholding comprehension.

So educators reach for the exception. And this is where it goes wrong in two directions at once.

The first failure mode is treating the exception as permission to check out entirely. The image gets alt="" or worse, alt="image" or alt="question 3 graphic". The rationale is understandable: “It’s a test, the exception applies, we can’t give away the answer.”

But the exception doesn’t say that. Read it carefully. 1.1.1 defines situations where the requirement doesn’t apply, including when non-text content is part of an assessment and the text alternative would fundamentally invalidate the assessment. The exception permits withholding the answer. It does not permit withholding comprehension.

The second failure mode comes from the accessibility side, oddly enough. An auditor or developer, terrified of a 1.1.1 failure, writes alt text that fully describes the content: “Five blue balls arranged in a circle.” Now the assessment is ruined for everyone, and the fix gets blamed on accessibility. This is the kind of thing that makes educators resent WCAG, and it’s hard to blame them.

That’s printing the answer key in Braille.

Here’s what Rob and I kept circling back to: the exception grants relief from equivalence, not from communication. The image still needs to be identifiable. The user still needs to understand what they’re being asked about, even when the tested skill requires vision.

Consider a counting question. The prompt is: “Which picture has 5 balls?” The answer options are four images. How do you label them?

The message is “nobody thought about you when building this.”

<img src="balls-1.png" alt="Option 1"> <img src="balls-2.png" alt="Option 2"> <img src="balls-3.png" alt="Option 3"> <img src="balls-4.png" alt="Option 4">

That’s not an exception. That’s abandonment. A non-visual user knows there are options and nothing else. They can’t even guess strategically.

<img src="balls-1.png" alt="Five blue balls"> <img src="balls-2.png" alt="Three yellow balls">

That’s not accessibility either. That’s printing the answer key in Braille.

The middle path describes the distractor characteristics without encoding the tested information:

<img src="balls-1.png" alt="Several blue balls"> <img src="balls-2.png" alt="Many yellow balls"> <img src="balls-3.png" alt="Some orange balls"> <img src="balls-4.png" alt="A handful of red balls">

Does this still require vision to answer? Yes. Counting is the tested skill; vague quantity words like “several” and “many” don’t leak exact counts. But the user now understands what each option depicts. They know the color palette, the general composition, the nature of the question they cannot fully answer. Compare that to “Option #” — which communicates nothing except that someone gave up.

Is that perfect accessibility? No. Let’s be honest about what this is: a documented, deliberate limitation with mitigations. The student using a screen reader still can’t independently complete this item. That means the real fix lives upstream in assessment design.

Offer an equivalent alternative item that tests counting non-visually (counting tones, counting tactile-equivalent descriptions, or differently structured questions). The exception buys you a compliant item, not an inclusive one. There’s a difference, and pretending otherwise is how we end up with “technically conformant” tests that still flunk real students.

While we’re here, the full set of situations in 1.1.1 that loosen the rules is worth having in one place:

Notice the pattern across all of them: every exception still demands you communicate something about purpose, or justifiably nothing at all. None of them grant permission for the lazy middle ground — present-but-meaningless.

When a student using assistive technology hits one of these images and hears “Option 2,” the message they receive isn’t “this item can’t be made accessible.” The message is “nobody thought about you when building this.”

Exceptions in WCAG exist because real pedagogy sometimes demands them. A music appreciation quiz needs you to hear the instrument. A reading assessment needs the passage unspoken. These are legitimate. But the exception is a scalpel, not a sledgehammer — use it to protect the integrity of the assessment, not as a blanket excuse to skip the thinking.

Every time you write alt="Option #", ask yourself: what could I tell this student without giving away the answer? The question “Which picture has 5 balls?” paired with “A handful of red balls” gives them dignity, context, and an honest shot at whatever strategy the item permits. Paired with “Option #,” it gives them nothing.