Writing

What we find when we open real webshops in a real browser. Method and raw numbers included every time — including the numbers that make our own product look less impressive, because those are the ones worth trusting.

31 July 2026 · research

We checked 27 Nordic webshops. None were clean.

One page from each of 27 Danish and Swedish shops, loaded in a real browser and audited the way a visitor's machine would. Every finding verified against the live DOM.

The result that matters is not the total. It is that four sites had a control a keyboard user simply cannot operate — and that axe-core alone found 92% of everything we found, which is the number that undercuts our own pitch.

  • 27 sites
  • 0 clean
  • 407 issues
  • 4 keyboard-inoperable

31 July 2026 · method

What axe-core cannot see

axe-core is good software and it runs inside our own tool. But it reads the accessibility tree, and some failures are invisible from there.

A <div> with an onclick and a pointer cursor looks and behaves like a button for anyone with a mouse, while being absent from the tab order and deaf to Enter and Space. Detecting that means comparing computed styles, event handlers and focusability together — which is a different question from the one a rules engine over the accessibility tree is asking.

Check one of your own pages

Free, no signup, about twenty seconds. A real browser opens the page and returns the findings with the code that fixes each one.

Run a check