How we test

What we check, how we check it, and what we can't check. Written for you, with a section at the end for your lawyer.

This is a technical accessibility assessment, not legal advice, and automated testing cannot establish ADA compliance.

The tool

We use axe-core, a free, open-source accessibility testing engine. It's the engine behind the accessibility checks in Google Lighthouse, which is built into Chrome. It checks your pages against the accessibility rules in WCAG 2.0, 2.1 and 2.2, Levels A and AA, that a computer can test. WCAG is the standard most ADA demand letters cite.

We add one check of our own: photo descriptions that are there but useless, like a file name (“IMG_4821.jpg”), a single generic word (“image”, “photo”) or a description over 150 characters.

Every report records the exact version of axe-core it used, so the same test can be repeated later.

Which pages we check

We check up to 15 pages of your site, picked in this order:

1. Pages you list when you start, and pages your letter names.
2. Your home page.
3. One page of each kind your site repeats, such as one product page and one collection page. Pages built from the same template usually share the same problems, so one of each covers a lot of ground.
4. A second page of each kind, then one-off pages like About or Contact, until we reach 15.

We find pages from the links on your home page and from your sitemap. On Shopify we also check the cart. We skip checkout, account and login pages, because they need a login or a payment.

Every re-check tests the same pages as your first assessment, so before and after compare like for like.

How each page is loaded

• We open each page in a real Chrome browser and wait for images and fonts to load before testing.
• If a page shows an error or a bot-protection screen instead of your content, we leave it out rather than report on the wrong page. Your report lists exactly which pages were checked.
• If your site is very slow, we stop after 10 minutes and report on the pages finished so far.

How we read your letter

We use AI to read your letter and pull out each problem it claims, quoted in the letter's own words so you can see exactly what it says. Your letter stays private and is never used to train any AI model.

The AI never decides what's wrong with your site. Only the scan does. We then match each claim to the issues the scan found. Claims we can't match go on your manual checklist.

How we set priorities

• High: your letter mentions it, or it's a serious problem of a kind these letters usually cite (photo descriptions, hard-to-read text, form fields, buttons and links with no name, page language) on a key page: home, product, collection, cart or contact.
• Medium: other problems axe-core rates as moderate or worse.
• Low: everything else.

What automated testing can't find

Automated tools find only part of accessibility problems. A computer can tell that a photo has no description, but not whether a description is accurate, or whether the whole site works with only a keyboard.

So every report includes a manual checklist: things axe-core flagged as needing a person's judgment, plus any letter claims we couldn't match. You can tick items off on your order page. Reports label those ticks “Reported by site owner. We didn't test these.”

Re-checks

After you fix things, a re-check tests the same pages again and gives each issue a status:

• Fixed (automated check): none found now.
• Improved: fewer than before.
• Unfixed: the same number or more.
• New: not found in the first assessment.

The best result we ever report is “0 detected by automated check.” We never tell you your site meets the ADA, because no automated test can show that.

For your lawyer: the evidence package

Each order includes a file called evidence.json. For every scan it records:

• When the scan started and finished, in UTC.
• The axe-core version used.
• The exact addresses of the pages scanned.
• Up to 3 screenshots of the problem elements for each kind of issue, each outlined in red, with the page it came from.
• A SHA-256 checksum of every screenshot and of the report PDF.
• The results of our report quality checks.

What the checksums are for: a checksum is a fingerprint of a file. Change one pixel of a screenshot or one character of the PDF and the fingerprint no longer matches. Anyone can recompute it with free tools, so your lawyer can show the files are the ones we produced on that date and haven't been edited since.

Before any report is released, we check it automatically: every page result is stored, no page is counted twice, the tool version is recorded, the report and PDF are complete and carry this disclaimer, and every screenshot matches its checksum. If any check fails, the report isn't released.