Overview
Accessibility testing is not one activity. It is three, and each one catches problems the other two miss. A store that only runs one type ends up with a false sense of security. The three types are automated testing, manual testing, and user testing with people who actually use assistive technology.
Automated testing
Automated tools scan your site’s code and flag common issues instantly: missing alt attributes, low color contrast, invalid HTML, missing form labels. Tools like Axe, WAVE, and Lighthouse fall into this category. It is fast, cheap, and consistent across every page at once.
The limitation is coverage. Automated tools can only reliably detect roughly a third of WCAG success criteria, since most failures require judgment a script cannot make, like whether alt text is actually descriptive or whether a heading structure makes logical sense. The full explanation is in why automated tools miss accessibility issues.
Manual testing
Manual testing is a person navigating your site with a keyboard only, and separately with a screen reader like NVDA, JAWS, or VoiceOver, checking each interactive element against WCAG success criteria one by one. This is where most real findings come from: keyboard traps, illogical focus order, form errors that are visible but never announced to assistive technology, buttons with no accessible name.
This is the core of what a real audit actually delivers, shown in what a real accessibility audit looks like.
User testing
User testing goes a step further: people who use assistive technology every day, not testers simulating it, attempt real tasks on your site, like completing a purchase or creating an account. This surfaces friction that technically passes WCAG but still creates a bad experience.
Most small and mid-size ecommerce audits do not include formal user testing due to cost, but automated and manual testing together still cover the large majority of what actually gets cited in lawsuits.
How they work together
Automated testing is best run continuously, after every theme or code change, to catch regressions fast. Manual testing is what you need before calling a store audited, since it is what actually verifies WCAG conformance. User testing is the highest-confidence layer, reserved for stores with the budget and traffic to justify it.
Example
A store ran only an automated scanner and passed, since the tool found zero contrast or alt text errors. A manual audit two months later found the checkout’s “Apply coupon” button had no accessible name and the cart quantity stepper was unusable by keyboard. Both would have gone into production undetected by the scanner alone.
Common mistake
Treating a clean automated scan as proof of compliance. A passing scan means the roughly 30 percent of criteria a machine can check are fine. It says nothing about the rest. If your budget only covers one type of testing, manual beats automated, since it is what actually catches the issues that end up in demand letters.