Zero axe-core Errors Isn't Compliance. Here's the Gap

Your accessibility vendor sends a report: zero automated errors. Your counsel asks the obvious follow-up: does that mean we are covered? The honest answer is no, and the distance between those two answers is where US ecommerce accessibility disputes tend to live.

A clean automated scan is valuable. It is also routinely misread as a certificate. This post explains what a zero from a tool like axe-core actually measures, what the 2026 data says about the automated floor, what no scanner can see, and how to describe your own results without overclaiming.

What a zero actually measures

axe-core is an open-source rules engine. It inspects the rendered page and flags failures that can be decided by machine: text that falls below a contrast ratio, form inputs with no programmatic label, images with no alt attribute, buttons with no accessible name, and similar rule-based checks. When a page returns zero violations, it means none of the rules the engine can evaluate were broken on that page, in that state, at the moment of the scan.

Each qualifier matters. "That page" means the other pages were not tested. "That state" means the cart drawer, the error message, the open modal and the logged-in account view may not have been scanned at all. "At the moment" means the next deploy can undo it.

A zero is a measurement. It is not a conformance claim.

The automated floor is not shrinking

The 2026 WebAIM Million analyzed the home pages of the top one million websites using February 2026 data, and the report was last updated on March 30, 2026. It found detected WCAG 2 failures on 95.9% of home pages, up from 94.8% in 2025, with an average of 56.1 detected errors per page, up 10.1% from 51 the year before.

The failures are the unglamorous kind. Low-contrast text appeared on 83.9% of home pages, missing alternative text on 53.1%, missing form input labels on 51%, empty links on 46.3% and empty buttons on 30.6%. WebAIM also reported that pages using ARIA averaged 59.1 errors, compared with 42 for pages without it, a reminder that adding accessibility attributes without testing them can make a page worse.

WebAIM states plainly that automated tools cannot detect every conformance failure, so these numbers are a floor. The practical reading for a US merchant: the most common failures are exactly the ones a scanner can find, and most sites still carry them. Clearing them is real, measurable progress. It is not the finish line.

How much does automation actually cover?

Vendors disagree on this, so treat any single figure with care. Deque, the company behind axe-core, published a study in March 2021 based on more than 2,000 audits covering over 13,000 pages and nearly 300,000 issues. It reported that on average 57 percent of accessibility issues by volume were covered by automated testing.

Two caveats from the study itself are worth keeping. First, that figure measures share of issues by volume, not share of WCAG success criteria, and a few high-frequency issue types such as contrast can pull a volume-based number up. Second, it is a vendor-published study. Even taken at face value, it leaves a large remainder that needs a human.

What no scanner can see

Rule-based checks can confirm that an image has alt text. They cannot judge whether the text is meaningful. They can confirm that a modal has a close button. They cannot confirm that a keyboard user can reach it, operate it, and return to where they were. The gaps that matter most in commerce are task-level:

• Can a keyboard-only shopper add a variant to the cart, change quantity, apply a code and reach payment?
• Does focus move logically through the cart drawer, and does it return to a sensible place when the drawer closes?
• Are checkout errors announced to a screen reader, and do they identify which field failed?
• Does the heading structure describe the page, or merely satisfy a rule?
• Do dynamic updates, such as price changes after a variant selection, reach assistive technology at all?

These are the journeys where a shopper either completes an order or abandons it, and they are the journeys a plaintiff's tester walks. No automated rule set covers them end to end.

Our own result, and the exact scope of the claim

MnT Future ran a WCAG 2.x AA audit on its own pages and took seven failing checks to zero, measured with axe-core, across nine key pages. The fixes were made in the design system rather than patched page by page, and no overlay widget was involved. Read the full WCAG 2.x AA compliance audit.

We describe that result precisely because the phrasing matters. It is an axe-core measurement across nine key pages. It shows that the machine-detectable failures we found were remediated at their source. It does not, by itself, establish full WCAG conformance, and we do not present it that way. A claim carries the instrument that produced it, or it should not be made.

That discipline is also the reason the design-system approach matters. Contrast, labels and focus styles defined once in shared components are fixed everywhere they render, including pages nobody scanned. A page-by-page patch fixes only the page you looked at.

How to build a record that holds up

For a US merchant or wholesale operator, a defensible accessibility record usually has four layers:

1. Source fixes. Correct the shared components and tokens, not individual templates.
2. Automated gates. Run axe-core in the build pipeline so a regression fails the deploy rather than reaching customers.
3. Manual task testing. Walk the revenue journeys with a keyboard and a screen reader on a schedule, and record the results.
4. Honest reporting. State what was tested, how, on which pages and when. Avoid words like "compliant" or "certified" unless a qualified assessor has issued that finding.

Layer four is the cheapest and the one most often skipped. Overstated claims are easy to check and hard to defend.

The short answer

Does a clean axe-core scan mean a site is WCAG compliant? No. Automated tools such as axe-core reliably catch machine-detectable failures like contrast ratios and missing form labels, but they cannot judge whether alt text is meaningful, whether focus order makes sense, or whether a task can be completed by keyboard and screen reader. A zero-error scan is a floor, not proof of conformance.

Where to go from here

If your accessibility report says zero and nobody has walked your checkout with a keyboard, you have a floor and no ceiling. A free strategy session with MnT Future's senior engineers can map which of your revenue journeys have been tested by a person, which have only been scanned, and what to fix at the component level first. See our commerce platform services or get in touch.