Site search
Runs the searches you set up and checks the result pages.
Limit: Other search words can reach other pages. You approve these searches, so they can send data to the site.
Coverage
What Axcess checks, and what it honestly can't
The Web Content Accessibility Guidelines (WCAG) 2.2 define 55 Level A and AA success criteria. Axcess contributes evidence to 29 of them today. This page shows where each result lands and how Axcess compares with Siteimprove and axe DevTools. It then lists all 55 criteria, with test steps for the 26 that a person must test by hand.
The checks
What each check does, and its limit
The same checks as the What was checked table in every report.
Opened in a browser
Opens each page in a real browser, so its scripts run and the browser checks can test it.
Limit: Opening a page is not a pass.
Rule check (axe)
Tests each page against the axe-core rules a computer can check, at the WCAG version and level you choose. Failures are Barriers.
Limit: A page with nothing found can still fail WCAG.
Rule check (Alfa)
Runs Siteimprove Alfa's standard test rules (Accessibility Conformance Testing, ACT). Each rule passes, fails, or cannot tell.
Limit: A failed rule is not proof the page fails WCAG. Each “cannot tell” goes to Needs review.
Image text check
Reads text in images (OCR) and asks a local vision model whether the alt text says the same.
Limit: A person confirms each result.
AI review
A local AI model reads link text, headings, labels, and instructions in context.
Limit: Results go to Needs review. A person confirms each one.
Keyboard check
Presses Tab, Shift+Tab, and Escape to find places keyboard focus gets stuck.
Limit: It only looks for traps. Test the rest of keyboard use yourself.
Zoom and layout check
Checks a 320-pixel-wide screen, 200% text, and wider text spacing for cut-off content.
Limit: A person confirms whether cut-off content is a barrier.
Click-Through
Opens menus, tabs, and dialogs, then runs the rule check (axe) on each page state it reveals.
Limit: It skips payments and other risky actions and blocks sending data, so a page state can be incomplete. It may not use every control it finds, and a dialog that would not close stops that page. A person checks custom controls and anything it did not reach.
Report groups
Every result lands in one of three groups
The group tells you how certain a result is and what to do next. The glossary defines each group.
Barrier
Confirm it, then fix it
Comes from axe-core rule failures, including problems found after clicking, and from Siteimprove Alfa failures. Confirm it on the page, fix it, then rescan.
Needs review
A person decides
Comes from browser checks, the keyboard check, text in images whose alt text is missing or does not match, local AI checks, and Alfa “cannot tell” results. Test it on the page and record your decision.
Informational
Nothing to fix
Comes from image checks where the alt text already matches, and from older records kept for history. No action is needed.
Compared with other tools
Where Axcess fits next to Siteimprove and axe DevTools
We use and like both. Siteimprove monitors whole sites from the cloud, and the axe DevTools browser extension checks the page in front of you. Axcess fills gaps: pages behind a sign-in, content behind a click, and checks that need a real browser or a person's judgement.
| Check | What it catches | WCAG | Report group | Also caught by Siteimprove? | Also caught by axe DevTools? |
|---|---|---|---|---|---|
| Rule engine (axe-core) | Machine-testable problems on every rendered page, such as missing text alternatives, unlabeled form fields, low text contrast, and ARIA errors. About 90 rules run at Level AA. | Many A and AA criteria, such as 1.1.1, 1.4.3, and 4.1.2 | Barrier | Yes. Siteimprove checks the same kinds of problems with its own rule engine. | Yes. axe DevTools runs these same axe-core rules. |
| Click through menus, tabs, and dialogs | Problems that only appear after a control is used. Axcess operates each safe control, runs axe-core again on what appears, and records which control revealed each problem. | The same criteria as axe-core | Barrier | Not automatically. Its crawler tests each page as it loads. A Dynamic Content Checker add-on lets you capture other states yourself. | Not automatically. axe-core does not test closed menus or dialogs, so you open them and scan again. Paid guided tests help with dialogs. |
| Reflow at phone width | Pages that need sideways scrolling at 320 CSS pixels wide. | 1.4.10 (AA) | Needs review | No automated check found. | No automated check found. |
| Text cut off at 200% zoom | Text that is clipped when the page is enlarged. Axcess approximates 200% zoom with a smaller browser window. | 1.4.4 (AA) | Needs review | Partly. It flags pages that block zooming. It has also documented a “Text is clipped when resized” rule, but its open-source engine deprecated that rule in 2026, so ask Siteimprove whether it still runs. | Partly. axe-core flags pages that block zooming, not clipped text. |
| Text spacing | Text that is clipped after line, letter, word, and paragraph spacing are raised to the WCAG values. | 1.4.12 (AA) | Needs review | Partly. It flags inline styles that lock spacing with !important. No check found that applies the spacing and looks for clipping. | Partly. The same inline-style check. |
| Keyboard traps | Places where Tab and Shift+Tab both fail to move focus away. Reported at most once per page. | 2.1.2 (A) | Needs review | No automated check found. | Only in the paid Keyboard guided test. No axe-core rule. |
| Focus checks | A focused control hidden behind a sticky or fixed header, footer, or banner, and elements that force the tab order with a positive tabindex. | 2.4.11 (AA), 2.4.3 (A) | Needs review | No automated check found. | Partly. axe-core flags positive tabindex as a best practice, and the paid Keyboard guided test reviews tab order. No rule found for focus hidden behind other content. |
| Target size | Buttons and links smaller than 24 by 24 CSS pixels without enough space around them. | 2.5.8 (AA) | Barrier | Yes, when WCAG 2.2 is selected. It also checks the stricter 44 pixel size (2.5.5, AAA), which Axcess checks only when Siteimprove Alfa runs at Level AAA. | Only when WCAG 2.2 rules are turned on. The axe-core rule is off by default, and Axcess turns it on for Level AA scans. |
| Text inside images | Images that contain words, found with OCR and compared with their alt text. An optional local vision model judges what the text is for and maps it to a WCAG criterion. | 1.4.5 (AA), 1.1.1 (A) | Needs review (Informational when the alt text already matches) | No automated images-of-text check found. Its open-source engine has an experimental rule that asks whether an image contains text. An on-demand AI check judges whether alt text matches the image. | No automated check found. A paid guided test reviews image alternatives. |
| Motion and reading order (optional) | Audio or video that plays on its own without controls, marquee text, and, with a local vision model, a visual reading order that differs from the code order. | 1.4.2, 2.2.2, 1.3.2 (A) | Needs review | Partly. Its open-source engine has an autoplaying audio rule that asks a person to confirm how long the audio plays, and a best-practice rule flags blink and marquee. No reading-order check found. | Partly. axe-core covers autoplaying audio (for review), blink, and marquee. No reading-order check found. |
| AI language checks (optional) | Vague link text, headings that do not describe their section, form fields without clear labels or instructions, and audio without a transcript, judged by a local AI model. | 2.4.4 (A), 2.4.6 (AA), 3.3.2 (A), 1.2.1 (A) | Needs review | Partly. It checks that links and fields have names, offers an opt-in AI rule for descriptive headings, and guides a review for transcripts. | Partly. axe-core checks that links and fields have names. Paid guided tests cover link purpose, headings, and labels. |
Pages behind a sign-in
Axcess scans them after you sign in yourself, including single sign-on and two-factor steps. Siteimprove's help center describes crawling behind a login through a setup done by its support team, and points to an add-on for sites that use multi-factor sign-in. Its free browser extension checks signed-in pages one at a time in your own browser, and so does the axe DevTools extension.
A second opinion from Siteimprove's engine
Axcess can also run Siteimprove Alfa, the open-source engine behind Siteimprove's checks, on your computer. Each result keeps the name of the engine that found it, so you can compare them. Results can differ from what the Siteimprove platform reports, which adds its own reviews and settings.
Checked in September 2026 against public Siteimprove and Deque documentation and the open-source axe-core and Alfa code. “No automated check found” means we found none in their documentation, not that none can exist. Tools change quickly, so please open an issue on GitHub if something here is out of date.
Manual testing
What you still need to test by hand
These checks still need a person. See Axcess and manual testing.
Keyboard
Use only the keyboard to reach and operate everything, with Tab, Shift+Tab, Enter, Space, the arrow keys, and Escape. Check that focus is always visible and moves in a sensible order. Axcess's keyboard check only looks for traps.
Screen reader
Listen to key pages and tasks with a screen reader such as NVDA, JAWS, or VoiceOver. Check that names, roles, headings, and announcements make sense in context.
Real tasks
Complete real tasks from start to finish, such as applying or registering, including error messages and time limits. Axcess does not submit forms or complete tasks on its own, apart from a search form you configure.
The criteria list below gives the manual steps for every WCAG success criterion. For guides and training, see LSA Accessibility's web accessibility resources and LSA Accessibility training.
At a glance
29 of 55 criteria with Axcess evidence
These numbers are generated from the coverage file the product itself reads, so they cannot drift from the code.
5 criteria
Axcess checks this reliably on its own. A person confirms it applies and reviews any remaining states.
18 criteria
Axcess catches the mechanical failures. A person tests the parts that need judgement.
6 criteria
A local AI model flags likely problems as leads. A person confirms each one before it counts.
26 criteria
Axcess has no check for this yet. The report gives you the steps to test it yourself.
| Level | Criteria | Automated | Partly automated | AI-assisted | Manual only | With Axcess evidence |
|---|---|---|---|---|---|---|
| Level A | 31 | 2 | 11 | 4 | 14 | 17 |
| Level AA | 24 | 3 | 7 | 2 | 12 | 12 |
| A + AA | 55 | 5 | 18 | 6 | 26 | 29 |
"Contributes evidence" is not "proves conformance". Even a fully automated criterion leaves a residual human judgement, which is why every card below has a What a person still checks section.
Explore
All 55 success criteria
Filter by how Axcess covers a criterion or by conformance level, or search by number, name, or keyword. Open any criterion to read what Axcess does and what a person still checks.
Coverage
Level
Showing all 55 success criteria
1. Perceivable 20 of 20 criteria
Can everyone perceive the content? Text alternatives for images, captions, contrast, zoom, and layouts that reflow on small screens.
1.1.1 Non-text Content
What Axcess does
axe flags missing alt on img / area / input[type=image] and unlabelled SVGs; the image-of-text VLM separately flags pictures that are really text.
What a person still checks
Whether the alt text that IS present is a meaningful equivalent, and the decorative-vs-informative call, needs a human.
Checks involved:Rule engine (axe-core)Image text checkConfidence: medium
1.2.1 Audio-only and Video-only (Prerecorded)
What Axcess does
For <audio> elements, the semantic LLM checks whether a transcript or text alternative is reachable (nearby "Transcript" link / surrounding text).
What a person still checks
Confirm the transcript is accurate and equivalent. Video-only (silent video) can't be detected from the DOM, still a manual check.
Checks involved:Meaning check (local AI)Confidence: medium
1.2.2 Captions (Prerecorded)
What Axcess does
No automated check yet. The report includes this criterion in the manual testing list with the steps below.
What a person still checks
Play each video and confirm synchronized, accurate captions. Auto-caption diffing (Whisper) is on the roadmap.
1.2.3 Audio Description or Media Alternative (Prerecorded)
What Axcess does
No automated check yet. The report includes this criterion in the manual testing list with the steps below.
What a person still checks
Confirm an audio description or full text alternative for prerecorded video.
1.2.4 Captions (Live)
What Axcess does
No automated check yet. The report includes this criterion in the manual testing list with the steps below.
What a person still checks
Confirm live audio in synchronized media has real-time captions.
1.2.5 Audio Description (Prerecorded)
What Axcess does
No automated check yet. The report includes this criterion in the manual testing list with the steps below.
What a person still checks
Confirm prerecorded video has a synchronized audio description track.
1.3.1 Info and Relationships
What Axcess does
axe checks list, table-header, definition-list, required-ARIA-children and heading-structure markup on the rendered DOM.
What a person still checks
Relationships conveyed only visually (grouped fields, columns, emphasis that implies meaning) need a human to confirm they're also programmatic.
Checks involved:Rule engine (axe-core)Confidence: medium
1.3.2 Meaningful Sequence
What Axcess does
The visual probe screenshots the page and asks a local vision model whether the visual reading order matches the DOM/source order (CSS can reorder content so screen readers hear a different sequence).
What a person still checks
Confirm the model's call by tabbing/reading with a screen reader. Subtle reorderings and content below the fold still need a human, and the probe only runs when a local vision model is available.
Checks involved:Visual and motion checkConfidence: medium
1.3.3 Sensory Characteristics
What Axcess does
No automated check yet. The report includes this criterion in the manual testing list with the steps below.
What a person still checks
Read instructions for reliance on shape/size/location/sound alone ("click the round button to the right"), judgement only a human can make.
1.3.4 Orientation
What Axcess does
No automated check yet. The report includes this criterion in the manual testing list with the steps below.
What a person still checks
Confirm content isn't locked to portrait or landscape (rotate the device / check for orientation-locking CSS).
1.3.5 Identify Input Purpose
What Axcess does
axe validates that any autocomplete tokens used are valid.
What a person still checks
Confirm autocomplete IS present on fields collecting the user's own info (name, email, address), missing autocomplete isn't auto-detected.
Checks involved:Rule engine (axe-core)Confidence: medium
1.4.1 Use of Color
What Axcess does
axe flags links distinguished from surrounding text by colour alone (a narrow heuristic).
What a person still checks
Most colour-only meaning, form errors, chart series, required-field markers, status, needs a human to confirm a non-colour cue exists.
Checks involved:Rule engine (axe-core)Confidence: low
1.4.2 Audio Control
What Axcess does
The visual probe measures actual playback advancement and flags audible audio longer than three seconds when no native or explicitly associated custom control is detected. Autoplay markup alone is not flagged.
What a person still checks
Confirm every audible autoplay source can be paused, stopped, or volume controlled independently; test custom controls and browser autoplay policy.
Checks involved:Visual and motion checkConfidence: medium
1.4.3 Contrast (Minimum)
What Axcess does
axe measures text/background contrast on the rendered DOM against the 4.5:1 (3:1 large-text) thresholds.
What a person still checks
Text baked into images, hover/focus/disabled states, and text over gradients or photos need a human to check.
Checks involved:Rule engine (axe-core)Confidence: medium
1.4.4 Resize Text
What Axcess does
The responsive probe zooms to a 200% proxy viewport and flags text that clips or overflows its container.
What a person still checks
Confirm no loss of content or function across the full zoom range in your target browsers.
Checks involved:Zoom and reflow checkConfidence: high
1.4.5 Images of Text
What Axcess does
OCR + a local vision model judge whether each image is really rendered text rather than a photo/diagram.
What a person still checks
Confirm flagged images aren't the allowed exceptions (logos, or text that's essential to a particular presentation).
Checks involved:Image text checkConfidence: medium
1.4.10 Reflow
What Axcess does
The responsive probe loads each page at 320 CSS px and flags horizontal scrolling / overflow.
What a person still checks
Confirm no content or functionality is lost in the reflowed view (some loss can pass the geometry check but still fail in use).
Checks involved:Zoom and reflow checkConfidence: high
1.4.11 Non-text Contrast
What Axcess does
No automated check yet. The report includes this criterion in the manual testing list with the steps below.
What a person still checks
Check that UI component boundaries (inputs, buttons, focus rings) and meaningful graphics meet 3:1. No reliable automated rule exists yet.
1.4.12 Text Spacing
What Axcess does
The responsive probe injects the WCAG text-spacing override CSS (line-height 1.5, etc.) and flags clipping/overlap.
What a person still checks
Confirm no text is cut off or overlapping with the spacing applied.
Checks involved:Zoom and reflow checkConfidence: high
1.4.13 Content on Hover or Focus
What Axcess does
No automated check yet. The report includes this criterion in the manual testing list with the steps below.
What a person still checks
For tooltips/popovers triggered by hover/focus, confirm they're dismissable, hoverable, and persistent.
2. Operable 20 of 20 criteria
Can everyone operate the site? Keyboard access, enough time, nothing that flashes, and clear ways to navigate and find things.
2.1.1 Keyboard
What Axcess does
axe flags some keyboard-inaccessible patterns; the keyboard probe confirms focus can move through the page.
What a person still checks
Confirm every control (menus, custom widgets, drag handles) is fully operable by keyboard, the deepest part of this SC is manual.
Checks involved:Rule engine (axe-core)Keyboard checkConfidence: medium
2.1.2 No Keyboard Trap
What Axcess does
The keyboard probe emits a review lead only when the same observable element resists repeated Tab and Shift+Tab exit attempts. It suppresses normal focus wrapping, small focus cycles, modal containment, and opaque iframe or closed-shadow focus.
What a person still checks
Reproduce every lead and test components that appear after interaction. Confirm whether arrow keys, Escape, a close control, or a documented non-standard command lets the user leave before recording a failure.
Checks involved:Keyboard checkConfidence: medium
2.1.4 Character Key Shortcuts
What Axcess does
No automated check yet. The report includes this criterion in the manual testing list with the steps below.
What a person still checks
If single-character shortcuts exist, confirm they can be turned off, remapped, or are active only on focus.
2.2.1 Timing Adjustable
What Axcess does
No automated check yet. The report includes this criterion in the manual testing list with the steps below.
What a person still checks
For any time limit, confirm it can be turned off, adjusted, or extended.
2.2.2 Pause, Stop, Hide
What Axcess does
The visual probe measures actual playback advancement for visible video longer than five seconds without a detected control. It also records <marquee> as an expert-review lead.
What a person still checks
CSS animations, auto-advancing carousels, and auto-updating regions aren't auto-detected, confirm any content that moves >5s can be paused, stopped, or hidden.
Checks involved:Visual and motion checkConfidence: medium
2.3.1 Three Flashes or Below Threshold
What Axcess does
No automated check yet. The report includes this criterion in the manual testing list with the steps below.
What a person still checks
Confirm nothing flashes more than three times per second. Flash analysis is not implemented.
2.4.1 Bypass Blocks
What Axcess does
axe checks for a skip link, landmark regions, and a heading structure that lets users bypass repeated content.
What a person still checks
Confirm the skip link actually moves focus and works with the keyboard.
Checks involved:Rule engine (axe-core)Confidence: medium
2.4.2 Page Titled
What Axcess does
axe checks that every page has a non-empty <title>.
What a person still checks
Confirm the title is descriptive and distinguishes the page (a light human check).
Checks involved:Rule engine (axe-core)Confidence: high
2.4.3 Focus Order
What Axcess does
The focus probe flags positive tabindex (WCAG failure F44), a manual tab order that overrides the natural DOM order and usually breaks the sequence.
What a person still checks
Tab through the whole page and confirm the focus order preserves meaning and operability, the order can break without a positive tabindex (e.g. CSS-reordered columns), which still needs a human.
Checks involved:Focus checkConfidence: medium
2.4.4 Link Purpose (In Context)
What Axcess does
axe flags empty/unnamed links; the semantic LLM judges whether the link text plus its context conveys where it goes.
What a person still checks
Confirm the LLM's borderline calls ("read more", icon links), it flags strong leads, not verdicts.
Checks involved:Rule engine (axe-core)Meaning check (local AI)Confidence: medium
2.4.5 Multiple Ways
What Axcess does
No automated check yet. The report includes this criterion in the manual testing list with the steps below.
What a person still checks
Confirm at least two ways to find pages (nav + search, or sitemap), except for steps in a process.
2.4.6 Headings and Labels
What Axcess does
axe flags empty headings / unlabelled controls; the semantic LLM judges whether each heading actually describes the content it introduces.
What a person still checks
Confirm the LLM's borderline heading calls, and check that form-control LABELS are descriptive, label descriptiveness is not yet AI-assisted.
Checks involved:Rule engine (axe-core)Meaning check (local AI)Confidence: medium
2.4.7 Focus Visible
What Axcess does
axe has limited checks for suppressed focus indicators.
What a person still checks
Tab the whole page and confirm a clearly visible focus indicator on every interactive element, largely manual.
Checks involved:Rule engine (axe-core)Confidence: low
2.4.11 Focus Not Obscured (Minimum)
What Axcess does
The live-page focus probe focuses each element and flags any whose centre is covered by a position:fixed/sticky overlay (the classic "focus hidden behind the sticky header" failure).
What a person still checks
Confirm partial-overlap cases the centre-point check can miss, and tab through interactively, overlays that appear only after a click still need a human.
Checks involved:Focus checkConfidence: medium
2.5.1 Pointer Gestures
What Axcess does
No automated check yet. The report includes this criterion in the manual testing list with the steps below.
What a person still checks
For any multipoint/path gesture (swipe, pinch), confirm a single-pointer alternative exists.
2.5.2 Pointer Cancellation
What Axcess does
No automated check yet. The report includes this criterion in the manual testing list with the steps below.
What a person still checks
Confirm actions fire on the up-event and can be aborted (no critical action on down-press).
2.5.3 Label in Name
What Axcess does
axe flags controls whose accessible name doesn't contain the visible label text (label-content-name-mismatch).
What a person still checks
Confirm the visible text is fully contained in the accessible name for voice-control users.
Checks involved:Rule engine (axe-core)Confidence: medium
2.5.4 Motion Actuation
What Axcess does
No automated check yet. The report includes this criterion in the manual testing list with the steps below.
What a person still checks
If a function is triggered by device motion (shake/tilt), confirm a UI alternative and a way to disable motion actuation.
2.5.7 Dragging Movements
What Axcess does
No automated check yet. The report includes this criterion in the manual testing list with the steps below.
What a person still checks
For any drag operation (sliders, reorder, kanban), confirm a single-pointer alternative (tap/click) exists.
2.5.8 Target Size (Minimum)
What Axcess does
axe checks interactive targets are at least 24x24 CSS px (with spacing).
What a person still checks
Confirm the inline / essential / equivalent-control exceptions are genuinely met for any flagged small targets.
Checks involved:Rule engine (axe-core)Confidence: medium
3. Understandable 13 of 13 criteria
Can everyone understand it? Readable language, predictable behaviour, and forms that help people avoid and fix mistakes.
3.1.1 Language of Page
What Axcess does
axe checks <html> has a present and valid lang attribute.
What a person still checks
Confirm the declared language actually matches the page's main content.
Checks involved:Rule engine (axe-core)Confidence: high
3.1.2 Language of Parts
What Axcess does
axe validates lang attributes that are present on parts of the page.
What a person still checks
Detecting foreign-language passages that are *missing* a lang attribute needs a human reader.
Checks involved:Rule engine (axe-core)Confidence: low
3.2.1 On Focus
What Axcess does
No automated check yet. The report includes this criterion in the manual testing list with the steps below.
What a person still checks
Confirm moving focus to a control doesn't trigger an unexpected context change (auto-submit, new window).
3.2.2 On Input
What Axcess does
No automated check yet. The report includes this criterion in the manual testing list with the steps below.
What a person still checks
Confirm changing a setting (select, checkbox) doesn't auto-trigger a context change without warning.
3.2.3 Consistent Navigation
What Axcess does
No automated check yet. The report includes this criterion in the manual testing list with the steps below.
What a person still checks
Confirm navigation repeated across pages stays in the same relative order. A cross-page embedding analyzer is on the roadmap.
3.2.4 Consistent Identification
What Axcess does
No automated check yet. The report includes this criterion in the manual testing list with the steps below.
What a person still checks
Confirm components with the same function are labelled consistently across pages. A cross-page analyzer is on the roadmap.
3.2.6 Consistent Help
What Axcess does
No automated check yet. The report includes this criterion in the manual testing list with the steps below.
What a person still checks
Confirm help mechanisms (contact, self-help) appear in the same relative order on every page that has them.
3.3.1 Error Identification
What Axcess does
No automated check yet. The report includes this criterion in the manual testing list with the steps below.
What a person still checks
Submit forms with invalid data and confirm errors are identified in text. Requires interaction the crawler doesn't perform.
3.3.2 Labels or Instructions
What Axcess does
axe checks a programmatic label exists; the semantic LLM judges whether each control's label/instructions are sufficient to know what to enter.
What a person still checks
Confirm the LLM's sufficiency calls, and test real form submissions, error-time instructions (SC 3.3.x) still need a human.
Checks involved:Rule engine (axe-core)Meaning check (local AI)Confidence: medium
3.3.3 Error Suggestion
What Axcess does
No automated check yet. The report includes this criterion in the manual testing list with the steps below.
What a person still checks
Trigger validation errors and confirm the page suggests how to fix them.
3.3.4 Error Prevention (Legal, Financial, Data)
What Axcess does
No automated check yet. The report includes this criterion in the manual testing list with the steps below.
What a person still checks
For legal/financial/data submissions, confirm reversal, checking, or confirmation is available.
3.3.7 Redundant Entry
What Axcess does
No automated check yet. The report includes this criterion in the manual testing list with the steps below.
What a person still checks
In multi-step flows, confirm previously-entered info is auto-populated or selectable rather than re-typed.
3.3.8 Accessible Authentication (Minimum)
What Axcess does
No automated check yet. The report includes this criterion in the manual testing list with the steps below.
What a person still checks
Manually test each in-scope sign-in and MFA step. Confirm it does not require a cognitive function test (for example, solving a puzzle or memorizing/transcribing information) without an accessible alternative. A successful post-MFA crawl only proves that an auditor established a temporary browser session; it does not automatically evaluate or pass the authentication experience. Do not record passwords, OTPs, passkeys, recovery codes, cookies, or session details in this report.
4. Robust 2 of 2 criteria
Does it work with assistive technology? Names, roles, and status messages that screen readers can rely on.
4.1.2 Name, Role, Value
What Axcess does
axe checks names/roles/values for standard controls and ARIA widgets (button-name, link-name, aria-* validity, roles).
What a person still checks
Custom widgets' state changes (expanded, selected, checked) need a screen reader to confirm they're announced.
Checks involved:Rule engine (axe-core)Confidence: medium
4.1.3 Status Messages
What Axcess does
axe checks for some live-region / role=status markup.
What a person still checks
Confirm dynamic updates (added-to-cart, validation, search counts) are actually announced, needs screen-reader testing.
Checks involved:Rule engine (axe-core)Confidence: low
Roadmap
Next on the list
Criteria the project has designed a check for but not yet shipped. They stay labelled manual until the code exists and has been validated.
3.2.3 Consistent Navigation
Embed each page's nav, cluster; outliers = nav diverges.
1.2.4 Captions (prerecorded video)
Transcribe the audio track, diff vs the published <track>; flag missing / garbage / desynced captions.
3.2.4 Consistent Identification
Find visually similar components across pages; verify their accessible names match.
3.2.6 Consistent Help
Locate help affordances across the crawl; flag pages where help moves to a non-standard location.
Roadmap and coverage are reconciled against the code in the coverage tracker.