Separate the position from the character
A key can reach the browser correctly while producing a different character from its printed legend. The test’s keyboard drawing is only a reference: choosing AZERTY here does not change your operating system.
Compare one ordinary key at a time
Write down the mismatch
Choose an ordinary letter or symbol, without a shortcut. Note what is printed on the physical key and what appears in a blank local document. Do not use passwords or personal text.
Check the selected input layout
Open your system’s input-language selector. Compare the exact layout with the physical keyboard: French and Belgian AZERTY differ. Keep any accessibility settings you rely on. Windows explains how to manage layouts in its input-layout guide.
Observe position and character together
Choose the closest reference in the test, start listening and press the key once. Read both Last key received and the position below it. The position called KeyQ can produce “a” with an AZERTY input layout. That difference alone does not mean a switch is faulty.
Change one setting and repeat
If the system layout is wrong, select the matching layout and repeat the same key in the test and your document. If only one app differs, review its shortcuts or remapping. If the position never arrives, switch to the guided one-key check.
Example: a Q label, an A character
Illustrative browser scenario, not a physical keyboard test. The browser reports KeyQ and the character “a”. With the US reference, the drawing labels that position Q; with the French reference, it labels it A. Changing the drawing changes neither the event received nor the system layout.
For Belgian Windows keyboards, compare the symbols as well as the letters. For instance, the reference for the 6 key differs from French AZERTY. This page uses Microsoft’s Belgian French reference, not a certified Mac mapping.
Dead keys and shortcuts need context
An accent key may wait for the next letter before producing a character. Fn, system shortcuts, accessibility tools and input methods can also change what reaches the browser. Compare in a local document; do not diagnose hardware from a printed label alone.
The report keeps observed positions and counts, not the characters you typed. Add the expected/received example yourself only if it is safe to share.