The denominator is what makes this honest. An engine that rate-limited us produced no measurement and is excluded; an engine that answered and named nobody produced a real finding and counts. Collapsing those two makes every throttled week look like a collapse in visibility.
It is reported per engine as well as overall, because the engines disagree with each other more than they agree, and a single blended percentage hides the disagreement that is usually the interesting part.
Most often confused with Share of voice, which is a different measurement with a different denominator.
Try it on your own site, free
No card, and the free tier does not expire — the first verdict takes about six weeks to arrive.