Click-through rate is clicks divided by impressions. A thousand impressions and fifty clicks is 5%. On an ordinary account it is one signal among many, and a low one is a reason to look at the ads. On a grant account it is the number the programme itself watches.
Google asks a grant account to keep an average click-through rate of at least 5%, measured across the ACCOUNT and over a month rather than per campaign or per day. One weak campaign does not fail an account on its own; it drags the average that does.
The two ways it goes wrong
Too broad is the common one: keywords that match searches the ad was never written for, collecting impressions from people who were never going to click. Too narrow is the quieter one — a handful of keywords so specific that the rate looks excellent on almost no traffic, which leaves most of the month's budget unspent and the mission unserved.
Both are visible in the same report. Read the search terms the account actually matched, not the keywords you entered: the gap between the two is where the rate is being spent.
Google's own wording: an account that misses 5% for two consecutive months can be temporarily deactivated, and is reinstated by request once the cause is fixed. Accounts using only Smart campaigns are outside the requirement. What happens to any particular account is still Google's alone to say.
Questions people ask
Per campaign or per account?
Per account, averaged over the month. That is why a single experiment left running can matter more than its own size suggests.
Is a very high rate good news?
Not on its own. Check what it was earned on: a 20% rate on four hundred impressions is a brand-name keyword doing well and the rest of the account doing nothing.
What is the quickest honest lift?
Negatives. Removing searches the ad was never for takes impressions out of the denominator without taking a single click out of the numerator.