Earlier benchmark scores have been removed because they lacked supporting test records. These checks evaluate a user-agent tool’s output without claiming measured vendor rankings.
Separate raw input from interpretation
A browser-identification tool may show a raw user-agent string, parsed browser or operating-system labels, and additional information available in the current browser. Keep those separate in your test. A label inferred by a parser is not the original request header.
Use cases that expose assumptions
- Current browser: save the raw string and compare it with the header received by an endpoint you control.
- Unknown input: if the tool accepts a custom string, submit an intentionally unrecognisable test value. It should preserve the input and avoid inventing a precise device identity.
- Mobile observation: record the actual browser and platform alongside the parsed result. Keep a real device observation distinct from a desktop browser’s emulation mode.
- Reduced detail: inspect whether unavailable version or platform details remain unknown rather than being guessed.
These are proposed tests. They do not state that WebsiteDown or a competitor has passed them. For a bug report, store the tool version or page URL and the time of each attempt.
Compare the same kind of observation
If a tool reports navigator.userAgent, compare that browser value with the actual outgoing header and record any difference. If it adds client-hint information, identify those fields separately. Comparing a parsed label with a raw header as if they were the same data can create a false disagreement. MDN documents the raw userAgent property and the limits of browser identification from the string.
Choose for reproducible bug reports
A useful tool preserves the raw value, marks uncertainty and makes the output easy to copy. Do not depend on browser identification alone to explain a rendering bug: capture the failing action, browser error and affected feature too. The presence of a browser name does not prove that every feature behaves like another installation of that browser.
Try the WebsiteDown user-agent checker with these tasks. We operate it; this is a first-party evaluation procedure. For a page that appears broken only in one browser, the session comparison checklist helps preserve the original failure.