Store owner’s guide / October 30, 2026

A Shopify accessibility checklist for the shopping journey.

By Sergio Fandiño · Storecessible

Start with one ordinary route: open the menu, find a product, choose an option, and review the cart. The goal is to spot a reproducible difficulty and describe it clearly enough for someone to investigate.

This is a limited first check, not a full accessibility audit or a test of legal compliance. Shopify’s theme accessibility guidance and W3C’s Easy Checks explain why a broader review remains necessary.

Choose a safe sample.

Use your own store or an authorized preview. Note the exact URLs, date, browser, window size, and selected product option. Avoid customer information and payment steps. If adding an item could reserve stock, run that part in a test store or preview instead.

1. Follow the path with a keyboard.

   Set the mouse aside. Use Tab and Shift+Tab to move between controls. Try Enter or Space where appropriate. Check that you can see the focus, reach important actions, and leave each component. A control that works only when hovered needs investigation.

   Record: the last reachable control and the next step you could not complete.

2. Open and close the menu.

   Try the menu trigger from the keyboard. Check its links, then close it. For a modal or drawer, test Escape and where focus returns. Compare the visible menu choices with the names exposed in the browser’s accessibility tools.

   Try our example: the improved synthetic demo lets you open its menu and return to the trigger with Escape. The separately labeled before version deliberately behaves differently.

3. Understand the product choices.

   Check the labels for size, color, and quantity. Confirm that a selected option is identifiable without relying on color alone. An image linked to a collection or product needs a useful link name, even when a neighboring text link points to the same place.

   Record: the option you selected and the visible and accessible names you encountered.

4. Check form instructions and errors.

   Look at search, newsletter, and other relevant fields without sending a real submission. A non-empty label can still describe the wrong task. Where you control a test form, check that an error explains what to correct and is associated with the field.

   Try our example: set the demo quantity to zero. Its improved version shows the error and focuses the quantity field; no order is created.

5. Check cart feedback with assistive technology.

   In an authorized test, add an item and check the confirmation, quantity, and remove controls. Test cart announcements with an actual screen reader. Finding a live region in the code alone does not prove that its message is spoken at the right moment.

   Record: the reader and browser versions, the action, and the actual announcement. If no reader was used, write “not tested”.

6. Inspect text and zoom.

   Check ordinary text contrast with a suitable measurement tool, considering size, weight, backgrounds, and transparency. Zoom the page and look for overlapping content or hidden actions. A narrow browser window and a physical phone are different test environments; identify which you used.

   Record: the exact element, settings, and measurement rather than a screenshot alone.

Turn an observation into a useful finding.

Use this compact record for each problem. Copy it into your notes or print this page from your browser.

Page and state

Exact URL, selected variant, open menu, or other relevant state.

Steps

What to press or choose to reproduce the problem.

Observed and expected behavior

What happened, what you expected, and which task became difficult.

Environment and evidence

Date, browser, device or viewport, reader if used, and a screenshot or recording.

Ownership and status

Theme, content, app, or not yet determined; confirmed, not reproduced, or not tested.

Keep scans and conclusions separate.

An automatic tool can help locate candidates, but it does not understand every shopping task, image, label, or interaction. Reproduce each proposed finding, avoid double-counting one shared component, and state what you did not test. A score is not a certificate.