This article reports on an audience costs experiment embedded into a survey of the British public (N = 2235). We extend previous research into audience costs in three main directions. First, we provide clear and direct evidence that they exist for a second-order democratic power, the United Kingdom. Second, we show that the extent of audience costs varies, and at times substantially, along with features of the crisis situation that have not been examined empirically in this context before. In particular, the type of crisis strongly influences public reactions both to bluffing in general and to the wisdom or otherwise of escalating crises before backing down. While audience costs do appear to exist for a UK Prime Minister, he or she cannot inflate them by moving up the escalation chain. Rather, a limited use of force tends to mitigate these costs among the British public because it appears to them a more consistent strategy. Third, we show that public disapproval of a failed bluffing strategy is stronger among the more politically engaged, increasing the likelihood that audience costs will be paid at the ballot box.