Skip to main content
WCAGrules
Quick navigation

Guides · Testing and tools

An Exam Program Is Rarely One Product

Scheduling, delivery, proctoring and accommodations are usually four systems with four reports. The first question is which one you meant.

Last reviewed August 31, 2026

Ask a testing provider for their conformance report and you may get a list. One major assessment provider publishes separate reports for eleven named systems, including a browser-based test driver, a desktop application installed on Windows, an online proctoring system, a scheduling and payment interface, an accommodations request system, and two ordinary marketing websites.

That list is the point. A candidate books on one system, sits the exam on another, is watched by a third and applies for adjustments on a fourth. One of those is not web content at all, so WCAG does not reach it. Asking whether the exam is accessible is asking about four things at once, and the answer will be four different answers.

So an exam audit starts by naming the system, then names the question types inside it, then names the configuration those questions run under. That is three inventories, and this page is about building them.

Naming the System Before Anything Else

The estate divides fairly reliably, and each piece has a different owner and a different chance of already having been evaluated.

SystemUsually built byWeb content?What an audit can cover
Registration, scheduling and paymentThe provider, or your own siteYesA complete process, from search to confirmation
Candidate account and results portalThe providerYesSign-in, results, certificates and downloads
Browser-based test deliveryThe providerYesThe question types, the navigation and the review screen
Installed test driver or lockdown browserThe providerNoOutside WCAG and outside what we test
Online proctoring in the browserThe provider or a subcontractorYesThe setup checks, the camera prompts and the wait states
Native proctoring applicationThe providerNoOutside WCAG and outside what we test
Accommodations request systemThe provider or your institutionYesOften the most consequential form in the whole estate
Course or revision materialYour own authorsYesCovered by our LMS page instead
Illustrative assessment estate. One row per system, filled in before scoping.

The accommodations row deserves a second look. A candidate who needs extra time or a different format has to apply for it, usually through a web form, often while under time pressure themselves. If that form is the least accessible thing in the estate, the whole accommodations policy sits behind a door some people cannot open. It is also the form nobody thinks to put in scope.

The Question Type and Configuration Matrix

This is the artifact. Question types down the side, and the configurations that change how each one behaves. A real matrix has your question types in it, not these, and the point of building it is that it makes the size of the job visible.

Question typeWhat the interaction depends onThe usual finding
Multiple choiceRadio group semanticsOptions not grouped, so the question is never read with them
Multiple selectCheckbox group plus a limitThe limit is enforced silently
Numeric or short text entryA labelled inputFormat requirements stated only in a tooltip
EssayA rich text editorThe editor's own toolbar, which is a component library nobody audited
MatchingTwo lists and a relationshipBuilt as drag and drop, covered below
Ordering or rankingA sequence the candidate setsDrag only, which is the named failure case for dragging movements
Drag and place onto an imageA pointer path onto a targetRarely has any non-drag route at all
Hotspot on an imageClicking a regionNo keyboard route, and no text description of the regions
Chart or graph plottingPlacing points on axesThe hardest type in the list, and the one most likely to need an alternative question
Code or formula entryA specialised editorKeyboard traps, and syntax feedback that is never announced
Calculator or tool useAn in-exam widgetA separate small application inside the timed window
Illustrative question-type coverage matrix. Each cell is a state to reach, not a component to inspect.

Then the configuration axis, which multiplies everything above. Timed or untimed. Extended time granted or not. One question per page or all on one. Back navigation allowed or blocked. A review screen or none. Flag for review available or not. Question randomisation on or off. Calculator enabled or not. Each combination that changes the interface is a state, and the ones you did not open are unevaluated rather than fine.

The Test Exception Is Narrower Than People Use It

There is a real exception in the standard for test content, and it gets stretched a long way past what it says. Non-text Content puts it like this. If non-text content is a test or exercise that would be invalid if presented in text, then text alternatives at least provide descriptive identification.

Two conditions and a consequence, and all three get skipped. The content has to be a test or exercise, so an illustration sitting beside a question is not covered. Text has to be the thing that would invalidate it, which is true of a question testing chart reading and false of a decorative photograph in the stem. And the exception does not remove the requirement, it reduces it to descriptive identification, so something still has to be there.

It is worth knowing that this is one of six situations in that criterion, and five of the six reduce the requirement rather than removing it. The general shape of 1.1.1 is not provide alt text or claim an exemption, it is provide what the situation calls for, and the situations are specific.

The Timer Argument, Made Properly

A timed exam has a genuine claim on the essential exception in Timing Adjustable, because extra time really can change what an assessment measures. That claim is available and it has to be made rather than assumed.

Essential is defined with two halves. Removing the limit would have to fundamentally change the information or functionality, and the same information and functionality could not be achieved in another conforming way. For a speeded assessment, both are arguable. For an assessment where the time limit exists to schedule seats rather than to measure anything, the second half is much harder.

And the exception is per timer, not per platform. The countdown on the exam has it. The clock on the scheduling page does not, and neither does the session timeout on the results portal. Writing down which timers exist and which of them can make the argument is a twenty-minute job that saves an afternoon of debate later.

Where accommodations sit in all this

W3C's own explanation of the timing criterion discusses a third party adjusting test time, which is a fair description of how assessment accommodations actually work. That is a policy mechanism rather than a technical one, and it does not remove the criterion's questions from the interface. If a candidate has been granted extra time, the timer still has to be perceivable, the extended limit still has to be communicated, and the request form that granted it is still web content somebody has to be able to use.

Submitting Answers Is Covered by a Rule People Miss

Error Prevention (Legal, Financial, Data) has four triggering categories, and one of them is test responses. W3C's explanation carries a whole situation for testing applications, which surprises people who assume the criterion is about money.

It asks for at least one of three safeguards, and one is enough. The submission is reversible. The answers are checked for input errors with a chance to correct them. Or the submission is confirmed, with a review step and a chance to change or cancel. A review screen listing answered and unanswered questions, with a route back to each, satisfies it and is good assessment design anyway.

What the criterion does not ask for is any judgment about whether the answers are right. Checking for input errors means checking the input, not the answer, and nothing in accessibility work touches academic validity.

What This Evaluation Does Not Reach

  • Installed software. A lockdown browser or a native proctoring application is not web content. WCAG does not govern it, and native testing is not something we sell.
  • Accommodation decisions. Whether a candidate should get extra time is a policy question for the awarding body. Whether the form asking for it works is ours.
  • Exam validity and psychometrics. Whether an accessible alternative to a chart-plotting question measures the same thing is a subject-matter question, and it needs the people who wrote the assessment.
  • A vendor's individual reports. A published index tells you which systems have reports. It tells you nothing about what any of them say until somebody opens them, and that is a reading exercise for whoever is buying.

One honest limit

We audit and never repair, so this is not an offer to rebuild a question type or a test driver. Most of an assessment estate belongs to a provider, so the findings that are theirs stay in their own section of the report. Testing needs a practice exam or a sandbox with the question types in it, arranged in advance, and a live sitting is never a test environment. Scope it at contact or read what the online course and LMS audit covers.

Common questions

Our exam vendor sent one conformance report. Is that enough?
Ask which system it covers. One major provider publishes separate reports for eleven named systems, and one of those is a Windows desktop application rather than web content. A single report almost never covers scheduling, delivery, proctoring and accommodations together.
Do exam images need alt text?
Usually yes. The test exception applies only where the content is a test or exercise that would be invalid if presented in text, and even then it does not remove the requirement, it reduces it to descriptive identification. An illustration beside a question is not covered by it at all.
Can we keep a timer on a high-stakes exam?
Yes, where the essential exception genuinely applies, and that has two halves. Removing the limit would have to fundamentally change the activity, and the same result would have to be unachievable another conforming way. The exception belongs to that timer, not to every timer on the platform.
Does accessibility work say anything about accommodations?
Only about the interface. Whether a candidate is granted extra time is the awarding body's policy decision. Whether the request form works with a screen reader, and whether the extended limit is communicated in the exam, are questions an evaluation answers.
What about drag-and-drop question types?
They need a non-drag single-pointer route, and a keyboard route separately, because the dragging criterion is not satisfied by keyboard support. Ordering and matching questions are the two most common shapes, and they are close to the failure examples W3C publishes.

Sources

Keep reading

More on testing and tools

Reading about it is the cheap part.

Find out where your site actually stands. The free scan checks 10 pages in a real browser against all 90 supported automated rules, keeps its 27 best-practice checks separate from WCAG findings, and names the rule behind every finding. The full audit adds an expert review and a real blind screen-reader user. From $499, with the report in 5 business days on Rapid and 10 on Standard, and the clock starting at cleared payment.

Go somewhere useful

Find tools, resources and your workspace.

29 destinations