Keyboard testing can reveal blocked controls, invisible focus, surprising order, and traps. It is one focused check, not a substitute for testing with disabled people, assistive technologies, or the full range of input methods.

Finish with

Outcome

A reproducible keyboard test record with the page and browser, intended task, focus sequence, observed barrier, consequence, evidence, owner, and retest result.

What you need

  • A desktop or laptop keyboard and a browser supported by the site.
  • A safe test account or non-destructive path if the task includes forms or private areas.
  • The expected primary task and permission to inspect without making irreversible changes.
  • A worksheet and, if available, a second browser or tester for confirming unexpected behavior.

The observable problem

Mouse use can conceal that a menu never receives focus, a custom control responds only to a click, or a dialog traps focus without a working close action. A visible outline alone is not enough if the order is confusing or activation does nothing.

Record behavior rather than broad conclusions. “After the third Tab press, focus moves from Search to the footer and never reaches Menu” is reproducible. “Navigation fails accessibility” is too general. Add the task consequence, browser, viewport, and page state so the owner can investigate the same path.

Invented for instruction

Fictional worked example

Fictional worked example: Cedar Room is an invented venue site. A visitor wants to view class times. The reviewer loads the homepage in a desktop browser, dismisses no pop-ups, and starts at the address bar. This is a fictional demonstration, not a report about a real service.

Tab reaches the skip link, logo, and search field, then jumps to the footer. The Classes menu never receives focus. Shift+Tab returns through the same sequence. The reviewer records a blocked primary route, not merely “bad focus order.” A direct Classes link in the footer works, but that workaround does not erase the missing menu control.

After a developer replaces the click-only element with an operable button and restores a visible focus style, the reviewer repeats the original sequence. Enter opens the menu, arrow keys behave according to the component’s documented pattern, Escape closes it, and focus returns to the button. The row is marked retested, while broader screen-reader and mobile checks remain open.

Procedure

  1. Define one task and reset the state

    Write the page, browser, viewport, sign-in state, and task, such as “open class times from the homepage.” Reload at the top and avoid the mouse after starting. Use a safe environment when activation could submit, buy, delete, publish, or expose private information.

  2. Move forward and record focus order

    Press Tab one step at a time. Record each interactive element in sequence, including skip links, menus, fields, dialogs, and footer controls. At every step, confirm that a visible indicator identifies the focused element and is not clipped or hidden behind sticky content.

  3. Operate controls with expected keys

    Use Enter or Space as appropriate for native controls. For composite widgets, follow the interaction documented for that component instead of assuming every arrow-key pattern. Check that labels, states, errors, and expanded content are perceivable without hovering.

  4. Reverse, escape, and recover

    Use Shift+Tab to move backward. Open overlays or menus and confirm there is a keyboard method to close them, commonly Escape when the component pattern calls for it. Ensure focus returns to a sensible control and never becomes trapped in hidden or inactive content.

  5. Complete the task and retest precisely

    Follow the primary path to its safe stopping point. Record the exact step where behavior differs from expectation, plus consequence and evidence. After repair, repeat the same browser, task, sequence, and state, then test a nearby route for regression.

Reusable worksheet

Keyboard-only test record. Use one row per meaningful focus stop or barrier.
Page, browser, and taskStep and keyFocused elementVisible and operable resultObservation and consequenceOwner and retest
Homepage; desktop browser; open class times4; TabFooter privacy linkVisible but menu was skippedClasses menu never received focus, so primary route is blockedFront-end owner; retest after button repair
Blank test contextNumber and keyUse visible or accessible nameRecord indicator and activationSeparate behavior from interpretationName owner, build, and result

Decision rules

  • If focus disappears or is not visually identifiable, stop and record the exact preceding and following elements before requesting repair.
  • If a primary control cannot be reached or operated, treat the task as blocked even when an indirect keyboard workaround exists.
  • If focus enters a component and cannot leave through expected keyboard interaction, stop destructive testing and escalate the trap promptly.
  • If activation could cause an irreversible action, use a safe test environment or stop before activation and arrange an authorized test.
  • If behavior differs across browsers, preserve both results and investigate support rather than selecting the more favorable result.

Limitations and escalation

  • Keyboard-only testing does not reproduce screen-reader interaction, voice input, switch access, touch exploration, magnification, cognitive barriers, or every browser and operating-system combination.
  • Passing one task does not show that every route, validation state, dialog, authenticated page, or dynamic update is operable.
  • The method finds observable behavior but may not identify the code-level cause. Native semantics, scripts, CSS, focus management, and browser behavior can all contribute.

Review list

Check the work

  • The record names the exact page, browser, viewport, state, and intended task.
  • Forward and reverse focus order were observed without using the mouse.
  • Every tested focus stop had a visible indicator and an understandable label.
  • Controls could be activated, closed, and exited with expected keyboard interaction.
  • Barriers include reproducible steps, consequence, owner, and retest result.
  • The record states which assistive technologies and other input methods were not tested.

Review again when

  • Navigation or component code changes
  • A dialog, menu, form, or account flow is added
  • Focus styling changes
  • A supported browser updates substantially
  • A user reports a keyboard barrier