What a Real Accessibility Monitoring Alert Looks Like (and the Three That Are Noise)

The alert lands at 2 a.m.: 41 new accessibility issues on the product listing page. By morning someone has triaged them, and 38 are the same alt-text flag on decorative icons, two are a color pair that only fails at a zoom nobody uses, and one is real. That ratio is why teams stop reading their alerts.

The alert worth waking up for

A real alert names a specific user harm on a specific page. A checkout button with no accessible name. A form error that appears visually but is never announced. A focus order that traps keyboard users in a modal. These are failures a person hits while trying to give you money, and they map to a WCAG criterion you can name: 4.1.2 for the button, 4.1.3 for the error, 2.1.2 for the trap. The alert should carry the criterion, the element, and the page, or it is not an alert, it is a count.

Volume is the other half of the signal. One new failure on the checkout template matters more than forty on a blog archive, because the template ships the failure to every page that uses it. Monitoring that reports raw counts without weighting by template or by traffic will always bury the one that costs revenue under the forty that do not. Read our take on the standard that changed this year: WCAG 2.2: The Six New AA Criteria, and Which One Breaks Your Site.

The three kinds of noise

First, the decorative-image flag. Scanners cannot tell a meaningful product photo from a divider or a background flourish, so they flag every image missing alt text, including the ones that should have empty alt text. If your team has already decided an image is decorative, the alert is not a finding, it is a scanner that cannot read your intent. Suppress it at the element, not by ignoring the whole rule.

Second, the duplicate. The same missing label on a component used forty times is one defect, not forty. A monitor that fires forty alerts for one component teaches the team that alerts are spam. Dedupe by component and template, and report the blast radius once.

Third, the criterion nobody enforces. A AAA-only flag on a site targeting AA is not a failure, it is a different standard. Mixing levels in one feed guarantees the feed gets ignored. Keep AA alerts and AAA observations in separate channels, or the AAA noise will eat the AA signal.

How to tune the feed so it gets read

Start with the pages that take money: checkout, cart, account, and the product page. Alert on new failures there within the hour, and alert on everything else in a daily digest. Then give each alert an owner by template, not by page, so the person who fixes the header component sees all forty pages at once instead of forty separate tickets.

The test of a tuned feed is simple: when an alert fires, does someone act on it the same day? If alerts sit unread for a week, the feed is too noisy, and the fix is not more alerts, it is fewer, better ones. The filings data backs up why this matters; see 3,117 Website Lawsuits in 2025. Here Is What the Count Actually Means. A monitoring feed that cries wolf every night is how a real failure ships unnoticed.

Ownership is the part most teams skip. Give each alert an owner by template, not by page, so the person who fixes the header component sees all forty pages at once instead of forty separate tickets. Then hold a weekly review of the muted rules: every suppression should carry a reason and an expiry, because a suppression written for a component that has since been redesigned is a real failure hiding behind an old decision. The feed is a living document, and a suppression without a review date is a failure waiting for a redesign to expose it.

Frequently asked questions

How many accessibility alerts should a store expect per day?

A tuned feed on a mid-size storefront should produce a handful of real alerts a week, not dozens a night. Dozens a night almost always means the feed is reporting duplicates, decorative-image flags, or criteria outside your target level.

Should I alert on AAA criteria if my target is AA?

Keep them in a separate channel. Mixing AAA observations into an AA alert feed trains the team to ignore alerts, and the AA failures that carry legal risk get buried with them.

Why does my scanner flag images that should have no alt text?

Scanners cannot tell decorative images from meaningful ones. An image that is pure decoration should carry an empty alt attribute, and the scanner flag should be suppressed for that element once a human has confirmed it is decorative.

What is the first page I should monitor closely?

Checkout, then cart, then the product page. Those are the pages where an accessibility failure stops a purchase, and they are the pages a plaintiff tester checks first.