This Arabic-first interface QA kit turns RTL and bidirectional-text principles into a repeatable review. Use the same stable IDs in both language editions, record pass, fail, or not applicable with evidence, and keep the tested build in view. The score is not a market benchmark or product comparison; it only summarizes the scope your team actually tested on a named build.
- deterministic checks
- 40
- adversarial strings
- 60
- review domains
- 10
Download the working edition
Start with an explicit direction contract
Set language and direction at the document boundary, then test components in RTL and LTR from the same source. The DIR checks cover logical properties, DOM order, and breakpoints, avoiding scattered left/right patches that pass one screenshot and fail another.
DIR-01 through DIR-04
lang and dir, logical properties, reading order, and breakpoints.
Isolate IDs, phones, and currency
The BDI group places email, IDs, phone numbers, and prices inside Arabic and English sentences with punctuation and parentheses. Preserve source values and isolate at presentation time with bdi or dir; do not store hidden direction controls as a UI patch.
BDI-01 through BDI-04
IDs, phones, currency, and punctuation without reordering or value mutation.
Mirror icons by meaning
The decision tree starts with meaning: does the icon encode movement on the reading axis? Previous, next, and progress usually mirror. Logos, media controls, clocks, and phones do not. For maps or domain symbols, document the contextual choice and test both the accessible name and hit area.
ICO-01 through ICO-04
Documented selective mirroring, never scaleX on a container that also flips text.
Test type and native content
The TYP and PAR groups review vertical metrics, diacritics, wrapping, and zoom, then verify that Arabic and English preserve the same task, state, and action without literal translation or sentence fragments assembled in one language order.
TYP-01 through TYP-04
Font coverage, line height, wrapping, and zoom.
PAR-01 through PAR-04
Task, state, link, and terminology parity across locales.
Review forms and complex data
FOR covers labels, validation, prefixes, and focus; DAT tests table headers, scrolling, chart axes, legends, and tooltips. Do not automatically reverse accounting order or geographic axes; data meaning decides.
FOR-01 through FOR-04
Label/help/error order, input direction, focus, and long validation.
DAT-01 through DAT-04
Tables and charts that remain understandable by keyboard and screen reader.
Exercise overlays, accessibility, and responsive states
OVL, A11, and RSP cover dialogs, menus, toasts, focus order, accessible names, and contrast, then repeat at narrow widths, 200% zoom, and with long content. Visual alignment alone does not pass when reading order changes or the first action disappears.
OVL-01 through OVL-04
Overlay placement, focus trap, dismissal, and status announcement.
A11-01 through A11-04
DOM order, names, contrast, and reduced motion.
RSP-01 through RSP-04
Narrow width, zoom, keyboard, and print.
Score transparently
Each check is pass, fail, or not applicable. Score = pass ÷ (pass + fail) × 100; exclude N/A only with a reason. Also report executed, failed, and critical counts, and never compare scores across different scope, build, or device matrices. A critical failure blocks release regardless of the average.
Auditable evidence worksheet
Record build, device, browser, check ID, evidence link, owner, and retest date.