GatekeeperQA / Testing resources
READ THE EVIDENCE
How to read an accessibility report.
Start with what was tested, then decide what needs a fix and what needs a person. A scan result is one part of your assessment, not the whole verdict on a website.
Published · Updated · By GatekeeperQA
This web accessibility report example walks developers, QA testers and agencies through actual reports from our own demo fixture. Start with the evidence below, then try the interactive sample or download the illustrative sample PDF. The interactive sample and PDF use synthetic findings; they are not the recorded demo results below.
1. Check coverage before the score
Confirm the page URLs, scan date, completed pages and execution errors. A URL identifies a page, but not every state of that page: an open dialog, form error or expanded menu may need a separate test. GatekeeperQA’s public scanner checks the initially loaded page state.
INCOMPLETE is a coverage warning. If a page could not be assessed, zero reported findings does not mean it passed. Read the recorded error, resolve the cause where possible and repeat the scan. Keep completed-page evidence, but state what remains untested.
2. Read the quality gate in context
- PASS: the automated result met the configured policy. Next, confirm coverage and complete the relevant manual checks. It does not establish complete accessibility.
- WARN: read why the warning occurred. It may reflect findings below the failure threshold or an unresolved check. Investigate the evidence; do not automatically call every warning a confirmed defect.
- FAIL: inspect the threshold explanation and the findings that exceeded it. Prioritize the affected journey, fix the barrier and retest the same scope.
For example, a policy allowing zero critical findings fails when one is found. That is a release-policy decision. It is different from a manual assessment of all applicable requirements. Reviewer notes and tracking statuses do not rewrite the original automated result.
3. Inspect the finding, not just its severity
Use severity to help prioritize, then consider the barrier in the actual user journey. A problem that prevents submitting an important form may need urgent attention even when other issues have similar labels. Severity labels are not WCAG conformance levels.
Read the rule, page, selector, captured HTML and available measurements together. A selector locates a rendered element; it is not a source-file line number. Report counts may include the same element under more than one rule, so do not treat their sum as a count of unique affected elements.
A real demo finding, explained
These preserved reports were recorded on September 25, 2026 using GatekeeperQA CLI 0.1.0 against our local demo fixture. They contain no customer assessment. The temporary localhost address in the reports is not a public website to visit.
Open the actual before report · Open the actual retest report
- Finding:
button-name, marked critical, with one affected element. This identifies a missing accessible name; “critical” is the scanner's impact label, not a WCAG level. - Location:
#checkout. Find that element in the tested page state rather than searching for a source-file line number. - Captured evidence: the empty native button below. Read the HTML alongside the rendered control to understand its purpose.
<button id="checkout"></button>
Change made: we added visible text to the same button. This supplies an accessible name as well as a visible label.
<button id="checkout">Checkout</button>
Retest evidence: the first report has one critical finding and a FAIL result; the second has zero confirmed findings and a PASS result under the same policy. The policy fails on serious or critical findings. This demonstrates that this rule no longer detected the barrier in the fixture, not that every interaction was tested.
Still to verify manually: keyboard focus reaches the button and remains visible, Enter and Space perform the intended action, and a screen reader announces a useful name and role. These are follow-up steps, not recorded manual test results. See the complete fix-and-retest example.
4. Treat “needs manual review” as unresolved
An automated check may be unable to reach a decision, for example when text sits over an image. That is neither a confirmed defect nor a pass. Inspect the captured evidence and test the rendered page in the relevant state.
If a contrast finding includes measured foreground and background colors, a ratio and a required ratio, use those details to reproduce it. When measurements are unavailable, do not invent them. Record the background, text styling, method and outcome of your manual check.
Also plan keyboard, focus, zoom/reflow and assistive-technology testing appropriate to the feature. A scanner’s manual-review items are not an exhaustive manual test plan. W3C explains why knowledgeable human evaluation remains necessary.
5. Turn evidence into a reproducible ticket
Give the developer enough context to find the barrier and check the change:
- Page URL, date, browser and page state.
- Rule, affected element selector and HTML snippet when available.
- Steps to reproduce and the observed impact.
- Expected behavior, supporting evidence and relevant rule guidance.
- Suggested fix to investigate and an explicit retest step.
Keep credentials and personal data out of shared tickets. GatekeeperQA’s copy-issue text and CSV exports help with handoff; copying into a tracker is not a direct integration.
6. Verify the same scope after the fix
Repeat the scan on the same URLs and manually exercise the relevant states. A finding that is no longer detected may have been fixed, moved or removed from the tested state. Confirm that the element still exists and that its behavior works before marking the issue verified.
Keep the original report, the retest and reviewer evidence together. Use the accessibility retesting checklist and reusable retest note to record scope, remaining findings and manual outcomes. Download reports you need to retain beyond the account’s retention window.
A quick review before you share
- Does the report identify what was tested and what was not?
- Are confirmed findings separate from unresolved manual checks?
- Can another tester reproduce each reported barrier?
- Does each verified fix have dated retest evidence?
Next: follow the complete scan-to-retest workflow or review scanner coverage and usage limits.