Your Green Score Won't Stop an ADA Lawsuit

ClearPath Team · 2026-08-08 · 6 min read

You run a scan. Lighthouse comes back green. WAVE shows zero errors. You breathe out, close the tab, and move on with your day. Then a demand letter shows up anyway, citing keyboard traps, confusing error messages, and a custom dropdown that a screen reader user couldn't operate at all. None of that showed up in your scan. That's not a bug in the tool. It's the tool working exactly as designed, which is the problem.

What automated tools actually catch

Scanners like Lighthouse, axe, and WAVE are genuinely good at what they do. They read the DOM and check it against rules that can be verified with code alone. That includes things like missing alt attributes, insufficient color contrast ratios, missing form labels, duplicate IDs, and empty links or buttons. Automated scanners parse the DOM and evaluate rules programmatically, and are best for catching structural issues at scale like missing alt attributes, empty buttons, duplicate IDs, and insufficient color contrast ratios. Run one of these tools regularly and you'll knock out a real chunk of technical debt fast.

But "a real chunk" is doing a lot of work in that sentence. WCAG 2.1 Level AA includes 50 success criteria, and only roughly 30% of these criteria, about 15 to 16 of the 50, can be meaningfully tested by automated tools. The remaining 70% require human judgment to evaluate correctly. Different studies land on slightly different numbers depending on methodology. One analysis of axe-core against manual audits found it caught 22.6% of the manually reported issues, while a more advanced scanner layered on top found 62.8%, a real improvement over standard scanning, but still 37 points short of what a human auditor finds. Even the widely cited "57% by volume" figure from Deque's own research is skewed heavily by color contrast, which automation detects at near-perfect accuracy and which accounts for a disproportionate share of all recorded violations. Strip out contrast and the real number drops fast.

What automated tools reliably miss

The pattern in what gets missed isn't random. It's everything that requires judgment about meaning, not just structure:

• Alt text quality. A scanner sees an alt attribute and marks the image compliant, even if the text is empty, a filename, or unrelated to what's actually in the image.
• Focus order. Tools can confirm elements are focusable. They can't tell you whether tabbing through the page follows a logical, predictable path.
• Error messages. A form can announce "error" without ever saying what field is wrong or how to fix it. That passes automated checks and fails real users every time.
• Custom components. Dropdowns, modals, tabs, and date pickers built with divs and JavaScript often look fine visually but announce nothing useful to a screen reader unless ARIA roles and states are wired correctly.
• Reading order and context. Whether content makes sense when read aloud in sequence is a judgment call no crawler can make.

Industry write-ups describe this using different language, but they land in the same place: automated scanners mainly check code and structure, and are very good at finding common technical issues, but accessibility also depends on user experience, keyboard usability, cognitive clarity, and screen reader behavior, and how real users experience and interact with the product. These areas often require human judgment. While your team works through fixes like these, a widget like ClearPath can give visitors on-page tools such as text resizing, contrast adjustment, and pause-on-animation, which can ease some of the friction for some users, though it won't rewrite a confusing error message or fix a broken custom dropdown under the hood.

Why a clean scan isn't a legal shield

Website accessibility lawsuits haven't slowed down. 3,117 website accessibility lawsuits were filed in US federal courts in 2025, a 27% increase from 2024. And this is happening while most sites believe they're in reasonable shape: the WebAIM Million 2026 report found that 95.9% of the top one million homepages had detectable WCAG failures, averaging 56.1 errors per page. Regulators have also gotten sharper about overclaiming. Automated tool vendors and overlay products that promised near-total compliance through automation alone have drawn scrutiny, and at least one company's compliance claim was explicitly banned by the FTC in 2025. The lesson isn't that automation is worthless. It's that a scan result was never designed to be a certificate, and treating it like one is exactly where the legal exposure lives.

A manual pass a non-specialist can actually run

You don't need to be an accessibility engineer to catch a meaningful amount of what automation misses. Two checks take under an hour combined and surface the issues that show up in most demand letters.

1. Keyboard-only test. Put your mouse away. Use Tab, Shift+Tab, Enter, and the arrow keys to move through your homepage, your navigation menu, and one full form (like checkout or contact). Can you see where focus is at all times? Does the order make sense? Can you open and close menus, modals, and dropdowns without getting stuck?
2. Screen reader spot check. Turn on VoiceOver if you're on a Mac, or install NVDA for free on Windows. Navigate the same pages with your eyes closed or your monitor off. Listen for whether images are described, whether form fields announce their labels, and whether an error message actually tells you what went wrong.
3. Zoom to 200%. Check that text doesn't get clipped, overlap, or force horizontal scrolling.

Write down anything confusing, silent, or stuck. That list is worth more than another automated report.

Prioritizing what comes back

Once you've combined an automated scan with a manual pass, you'll likely have more findings than time. Sort by impact, not by the label the tool gave it:

• Fix anything that fully blocks a task first: keyboard traps, forms that can't be submitted, checkout flows that silently fail.
• Next, fix issues that repeat across every page, like a broken skip link or a nav menu that doesn't announce state.
• Then work through page-specific content issues: image descriptions, heading structure, individual form fields.
• Re-run both the automated scan and a quick keyboard pass after each round of fixes, since one change can affect focus order elsewhere.

A green score tells you that the easy 30 to 40 percent is handled. It says nothing about the rest, and the rest is usually where the actual complaints come from. Run the scan, then go do the keyboard test yourself. It takes fifteen minutes and it will show you more than another tool ever will. If you install something like ClearPath along the way, treat its generated statement and toolset as a helpful starting layer for visitors, not a substitute for the manual pass above.