This translation has not been editorially reviewed yet. The German version is authoritative. Deutsche Fassung
Responsive
How reliable are automated website checks?
An automated check is only as reliable as its protection against random findings. The audit service therefore collects every responsive finding twice per device size, with a short interval between. Only what both measurements agree on gets reported. If a result fluctuates, that's explicitly stated in the report — with the recommendation to look at the spot on a real device.
Web pages are in motion while loading
A website is in motion while it builds: images load in, fonts get swapped, reveal animations run, ad banners and embedded content appear with a delay. A single snapshot can catch the page in the middle of this motion — and then report an error that no longer exists a second later. An element briefly overhanging the edge, text that doesn't yet fit before the font swap: moments like these aren't shortcomings, they're intermediate steps of loading. Reporting them as findings sends customers hunting for errors that don't exist on the fully loaded page.
Two matching measurements support a finding
The sequence for every device size: the page loads, settles, gets measured — then the check waits briefly and measures again. For shortcomings, the cautious principle applies: only what both runs see gets counted. For the amount of text on the first screen, the higher of the two values counts — in doubt, in your website's favour, so a slow-building page gets treated fairly. If the two runs disagree on horizontal scrolling, the report claims nothing; instead, it notes that the finding fluctuates between two measurements, and recommends checking on a real device. That way, every statement in the report stays solid.
A false alarm weighs heavier than a missed finding
Behind the double measurement lies a basic stance of the measurement engine: a false alarm does double damage. It sends your agency hunting for an error that doesn't actually exist, and it devalues every other finding in the same report — once you've chased a phantom finding, you trust the real ones less too. That's why cautious rules apply throughout: only genuine horizontal scrollability counts as a scrollbar, collapsed menus never count as cut-off text, individual pixel-level edge cases are tolerated. At the same time, the opposite rule applies: an empty check is a failure, never a free pass. If a measurement fails, that's stated openly in the report, instead of appearing as a passed check.
- Read the measured-value column in the audit report — it states what the check actually saw.
- Take the note "verify on a real device" seriously, and open the affected page on a phone.
- For fluctuating findings, ask your agency about animations and content that loads in afterward — they're the most common cause.
- Compare two audit runs before commissioning major changes — a single outlier isn't a trend yet.
