“Not measured” is not a zero
A zero is a result. This is the absence of one. Anywhere those two get collapsed, a broken integration on our side starts looking like a competitor beating you.
The difference, in one line
Missing means the question was asked, an answer came back, and you were not in it. Not measured means no usable answer came back at all, so nothing whatsoever was established about you.
Those get printed as different things, counted as different things, and coloured as different things. The temptation to round the second into the first is real, because a result full of dashes looks worse than a result full of noughts. It is also the single easiest way for a measurement product to end up lying.
What actually produces one
Every not-measured run carries a reason, and the reason is shown to you. A gap with no explanation attached turns into a zero in the reader’s head in about a second.
- engine not configured
- That assistant has no key on this system. Its runs are marked before a single call goes out, so you can see they were planned and never issued.
- billing exhausted
- Our account with that provider has run out of credit. Nothing to do with your website. It is reported separately from a rate limit on purpose: this one never clears by waiting, and calling it a rate limit would hide an outage behind a routine-sounding word.
- rate limited
- The provider told us we were asking too quickly. Worth re-running later.
- timeout · engine error
- The call went out and nothing usable came back — too slow, or an error from their end.
- call cap reached
- Each scan carries a hard ceiling on how many calls it may spend. If it binds, the scan stops there, and every run it did not reach is recorded as not measured. Half a scan is never reported as if the other half had answered.
- homepage not usable
- Your homepage could not be read — a challenge screen, an error page, a page with nothing on it. No questions are written and no calls are made, and you get told what came back instead.
- your share of the day
- Five checks per internet connection per day, plus a ceiling on what the whole system may spend in one day. Both refusals carry not-measured and a reason, and neither returns a result. An office or a household behind one connection shares that count, which is the one case where this lands on somebody who has run nothing.
What it does to the numbers
- It is excluded, not counted. The denominator is measured runs, never attempted ones. A failed call is not allowed to drag a citation rate down below what was actually measured.
- A question with nothing measured is labelled unmeasured rather than printed as 0 of 0. Almost everybody reads 0 of 0 as a nought.
- It never counts against you anywhere. Not in the headline, not in the per-question table, not in the competitor tally.
The one that is not a queue
Google’s AI Overviews — the answers at the top of a Google results page — are not covered by this check, and not because nobody got round to it. Google publishes no way to ask for them. The only route anyone has is a paid vendor scraping the results page, and a scrape carries what a scrape carries: it changes when their layout changes, and it breaks quietly.
So that surface is named as not measurable rather than left off the engine list without comment. If it is ever covered here it arrives through a paid results-page vendor and says so on the face of the number, because a scraped figure and an answer from an engine’s own interface are not the same kind of fact and must not be printed as though they were. The rest of what is out of scope is written down too.
Why this rule is worth the ugly reports
A half-down check is the most tempting thing in this product to tidy up. Two assistants asked, one answered, and the honest version of that result says one engine of two — which reads worse than a clean scan, and that is exactly the point. The alternative is a number that looks fine and is not true, and there is no way to run a measurement business on the other side of that line.
If this page did not answer it
Write to the one inbox. There is no chat widget, no ticket queue and no phone number — a message arrives with the person who wrote the thing you are asking about, and gets answered by hand. Send the domain you checked, what came back, and what you expected instead. That difference is the whole message.
Any question that arrives twice becomes a page here, so writing in is also how this section gets longer.
What varies, what that makes a single check, and what is refused because of it.