KEY & CLICK · OPEN DOCUMENTATION
Sources & method
A browser can show which key events reached a page. That is useful evidence, but it is not a laboratory measurement or hardware certification. Reviewed 25 September 2026.
What the counters mean
Coverage counts distinct observed positions among the included keys in the current diagram. Fn is optional; manually excluded keys are listed separately. Hold repeats are counted separately from fresh presses. The highest simultaneous count describes this session, not guaranteed NKRO. The tool cannot measure hardware latency or USB polling rate.
Language, region and automatic suggestions
Only the root address chooses a language automatically: saved language, then a supported browser language, then a Russian regional hint, otherwise English. /en/ and /ru/ always honor their address. OS labels are inferred from browser information. Regional preference is independent of interface language. No IP lookup or precise location permission is used. An available keyboard map may suggest legends before testing starts; manual choices take priority. Shape and size remain reference choices, not hardware detection.
Where the diagrams come from
W3C defines the physical code identifiers. Microsoft references inform Russian, Brazilian and French legacy legends; the Japanese 106-key driver is the JIS punctuation reference. Apple documents differences among ANSI, ISO and Japanese hardware. Our 60%, 65%, 75% and laptop diagrams are generic references, not exact reproductions of every model. Apple punctuation and special keys can differ from PC references. Hardware testing on every model has not been performed.
Why English and Russian first
This is a staged localization choice, not a worldwide demand ranking. In Similarweb’s August 2026 desktop estimate for the benchmark key-test.ru, India represented 20.79%, Russia 14.92%, and Indonesia 12.65%. Russia was second for that site and scope. Competitor estimates are not our audience data, unique visitors, or a revenue guarantee. We will compare actual use before adding languages.
Test data and future measurement
Key events stay in this tab. Only preferences are saved on this device; key history and per-key counts are not saved between visits or automatically transmitted by this tool. User-requested reports and authorized read-only browser assistance can access the test summary; they omit recent character history. This prototype has no analytics or advertising integration. Future measurement should use coarse events such as test started, mode selected and report opened, without key codes, characters, combinations or event-level timestamps. One observed key may satisfy a visitor; it does not prove their problem is solved. We will not infer age or gender from typing. Hosting still receives ordinary page requests.
Readable status and accountable corrections
Symbols and written status accompany color. Untested and excluded lists make the coverage denominator visible. Sources are linked where they support a decision. Corrections distinguish changed facts from clarified limitations; a citation does not mean the source publisher endorses our implementation.
Mouse and gamepad observations
Mouse button numbers represent system-assigned roles, not a guaranteed physical left/right arrangement. We count received presses and browser double-click events only within the active test surface. Wheel input can originate from a mouse or trackpad. Gamepad readings come from the browser; the OS or driver may already have processed them. Only a standard mapping assigns control positions. Rest measurements use the largest radial distance observed over five visible, focused seconds; no universal fault threshold, dead-zone filter or recentering is applied. A gap longer than 250 milliseconds interrupts this observation; this is a browser-sampling continuity rule, not a hardware fault threshold. A hidden or unfocused page pauses testing. These tools do not measure sensor DPI, end-to-end latency or certified hardware polling rate. The linked MDN references explain the input APIs; Microsoft explains center variation and deadzones. Their publishers do not endorse this implementation.
Caps Lock evidence and live letter previews
MDN’s modifier-state reference documents what getModifierState reports and where browser or platform support differs. A false report alone cannot establish support. Our conservative indicator shows ON on a positive CapsLock report; it shows OFF only after this session has previously reported ON. Otherwise it shows ?. After focus loss we wait for a new event. The optional live view can also use a received letter to preview case without claiming that this proves Caps Lock is off. MDN’s key reference explains how modifiers, layout and composed input affect event.key. We compare received letters with the selected reference, apply layout-aware case pairs, and pause the preview for shortcuts, IME or a mismatch. This is a case preview, not an operating-system input-source detector or an exact typing emulator: remapping and unusual input methods can differ. The actual received value remains visible in key details. These documents support the method, not certification of this implementation.
What has been checked, and what remains device-specific
Automated checks cover session counting, capture policies, language preferences, layout suggestions, and reference-grid consistency. Local browser checks cover representative interactions and page sizes. These are software checks, not a physical certification of every keyboard. Mac toggle behavior, AltGr, Japanese IME keys, media keys, assistive technology and browser-reserved shortcuts still need comparison on the actual OS, browser and device. The test deliberately reports observations rather than a pass/fail hardware verdict.
Operating-system icon sources
The Windows logo, Apple logo used for macOS, and Tux penguin used for Linux come from Bootstrap Icons by the Bootstrap Authors. Each linked icon page provides its SVG and usage examples. We retain the original SVG files and MIT license notice locally; the selector uses their vector paths with colors matched to its state. These marks identify the choices and do not indicate sponsorship or endorsement.
Color roles and readable contrast
Our palette is a product design choice, not a universal hardware-test color standard. Green marks observed keys, strong blue marks keys held now, soft blue groups selections and recent observations, and amber invites review of unobserved keys. Amber is not a fault diagnosis. The W3C color guidance below explains why labels and symbols must accompany meaningful colors; its contrast guidance explains the 4.5:1 minimum for ordinary text. We measured the new blue and amber text combinations above 4.5:1. This targeted check does not claim whole-site accessibility certification.