WCAG checker vs accessibility audit: what each actually finds
Published: 7 min readThe AccessibleOrNot team
A WCAG checker is software that tests pages against the WCAG rules that can be expressed in code; an accessibility audit is an evaluation by a trained person who uses the site with a keyboard, screen readers and other assistive technology and judges every applicable WCAG success criterion. A checker finds about 57% of issues by volume, in minutes, on as many pages as you like (Deque, 2021). An audit covers all criteria, but on a sample of pages and at one point in time. Most sites need both: the checker for breadth and repetition, the audit for depth.
In short
- A WCAG checker tests rules that can be decided from code; an audit has a person judge every applicable WCAG success criterion.
- Automated testing found about 57% of issues by volume in Deque's study of more than 2,000 audits (March 2021); by criterion, the share that can be fully automated is much lower.
- Several WCAG 2.2 criteria, such as 2.5.7 Dragging Movements, 3.3.7 Redundant Entry and 3.3.8 Accessible Authentication, can't be judged by a checker at all.
- The W3C's evaluation method, WCAG-EM 2.0 (published 23 July 2026), describes an audit in five steps: scope, explore, sample, evaluate, report.
- An audit is a snapshot; a checker repeated weekly is what catches regressions after the audit.
What is the difference between a WCAG checker and an accessibility audit?
The difference is who decides. A checker applies fixed rules: "does this image have an alt attribute?", "is this contrast ratio below 4.5:1?". An auditor asks what the rule is for: "does this alt text tell a blind customer what the product looks like?", "can I finish checkout without a mouse?".
| WCAG checker | Accessibility audit | |
|---|---|---|
| Who decides | Rules in software (often axe-core) | A trained person, with assistive technology |
| WCAG coverage | Criteria that can be decided from code, fully or partly | Every applicable success criterion |
| Page coverage | As many pages as you scan | A representative sample |
| Time | Minutes per page set | Days, depending on scope |
| Repeatable | Yes, at no extra cost | Each repeat is a new audit |
| Output | Findings with element, location and rule | Findings per criterion with pass/fail, explanations, often a conformance report |
| Blind spots | Meaning, behaviour, flow | Pages outside the sample; changes after the audit |
Source: Deque Systems, W3C WAI
What does each one actually find?
The table below takes common issues on a shop or service site and says which method finds them. "Partly" means the checker finds the issue in some forms but not others, or flags it as "needs review" for a person to decide.
| Issue | WCAG | Checker | Audit |
|---|---|---|---|
| Image has no alt text | 1.1.1 | Finds it | Finds it |
| Alt text is "product-image-3" | 1.1.1 | Partly | Finds it |
| Grey price text at 2.3:1 contrast | 1.4.3 | Finds it | Finds it |
| White text over a photo banner | 1.4.3 | Partly (needs review) | Finds it |
| Icon button with no name | 4.1.2 | Finds it | Finds it |
| Mega menu won't open with a keyboard | 2.1.1 | Misses it | Finds it |
| Newsletter pop-up traps focus | 2.1.2 | Misses it | Finds it |
| "Added to cart" isn't announced to a screen reader | 4.1.3 | Misses it | Finds it |
| Error says "Invalid input" with no field or reason | 3.3.1, 3.3.3 | Misses it | Finds it |
| Colour swatch is 16×16 px | 2.5.8 | Partly | Finds it |
| Slider can only be dragged, no buttons | 2.5.7 | Misses it | Finds it |
| Checkout asks for the address twice | 3.3.7 | Misses it | Finds it |
| Login needs a puzzle CAPTCHA with no alternative | 3.3.8 | Misses it | Finds it |
| Same issue on 4,000 product pages | any | Finds every instance | Finds it on the sampled pages |
The last row is the checker's strength. An audit tells you the product template has a problem; a checker tells you on which of your 4,000 product pages it shows up, and whether it's still there next week.
Source: W3C, Deque Systems, GitHub
How does an expert accessibility audit work?
The W3C published version 2.0 of its WCAG Evaluation Methodology (WCAG-EM) on 23 July 2026. Version 1 was for websites; version 2 also covers apps and other digital products. It describes an audit in five steps:
- Define the scope: which site or app, which WCAG version and level (for example, WCAG 2.2 AA).
- Explore: find the key pages, functions and templates (search, product, cart, checkout, account).
- Select a sample: a representative set of pages and flows, including complete processes such as checkout.
- Evaluate: test each page in the sample against every applicable success criterion, with a keyboard, screen readers, zoom and code inspection. Automated tools are used here too, as an aid.
- Report: record the result per criterion and explain the failures. Many auditors also prepare a conformance report in the format procurement teams ask for (for example, based on the VPAT template).
Two things to ask an auditor: which assistive technology and browser combinations they test with, and whether a retest of fixed issues is included.
Source: W3C WAI
Where does each method fall short?
- A checker can't judge meaning or behaviour. A clean report means "no issues that can be found automatically", not "no issues". The W3C says evaluation tools can't determine accessibility; they can only assist.
- A checker can be noisy. Items marked "needs review" still need a person; skip them and you miss real problems.
- An audit is a snapshot. The day after the report, a new campaign banner or theme update can bring back a fixed issue. Nothing in the audit will tell you.
- An audit sees a sample. If a problem only appears on a template the sample didn't include, it stays hidden.
Which do you need, and when?
| Situation | Start with | Then |
|---|---|---|
| You've never tested the site | A checker on every template | A 20-minute manual test on your main flow |
| Before a redesign or new theme goes live | An audit of the key templates | A checker in the release process |
| A customer complaint or legal letter | An expert audit of the affected flow | Fixes, retest, a dated record |
| A buyer asks for a conformance report | An expert audit | A checker to keep the report true |
| Small team, small budget | A checker plus a manual test | An audit of checkout once the volume is fixed |
| Ongoing | A checker every week | An audit every year or after major changes |
The order that saves the most money: fix what the checker finds first, then bring in the auditor. Auditors' time is the expensive part; don't spend it listing missing alt text a tool could have found for free.
Frequently asked questions
Can a WCAG checker replace an accessibility audit?
No. A checker finds about 57% of issues by volume (Deque, 2021) and can't judge meaning, keyboard behaviour or flow. An audit judges every applicable WCAG criterion.
Is an automated scan an audit?
We don't call it one. In our terms, a scan is automated and an audit is done by a person. Some vendors call an automated report an "audit"; ask who reviewed it and which criteria were judged by hand.
How long does an accessibility audit take?
It depends on the scope and the sample. WCAG-EM 2.0 doesn't set a duration; ask the auditor how many pages and flows they'll test and whether a retest is included.
Does the EAA or the ADA require a third-party audit?
Neither names a mandatory third-party audit for websites. An audit is still the strongest evidence of the work you've done, alongside a dated record of fixes and monitoring. This isn't legal advice.
Sources
- Automated testing study identifies 57 percent of digital accessibility issues · Deque Systems · 10 March 2021
- WCAG Evaluation Methodology (WCAG-EM) 2.0 — Note Published · W3C WAI · 23 July 2026
- Selecting Web Accessibility Evaluation Tools · W3C WAI · 13 May 2024 (updated)
- WCAG 2.2 · W3C · 5 October 2023 (current version)
- axe-core rule descriptions · Deque Systems, GitHub · current
Related articles
Clear the checker's share before the auditor arrives.
Enter your URL. We scan your templates, list every automated finding with its location and a code example, and give you a manual checklist for what only a person can judge.