Estimating changes in the frequency or height of extreme sea levels (ESLs; e.g., the 100-yr event) is a popular approach for illustrating future coastal flood risk to societies under various climate change scenarios. However, these metrics only account for physical water levels (i.e., the hazard). They do not consider societal outcomes (e.g., loss of life, property damage). As a result, physical ESL metrics and associated thresholds may give misleading estimates of future coastal flood risk. This has implications for climate adaptation decision-making and risk communication efforts that seek to quantify changes in coastal flood risk under different climate scenarios. Here, we illustrate how some risk measures can lead to sizable differences in estimates of future coastal flood risk, relative to when only considering physical impacts by considering 1) projected ESLs under +2 degree C and +5 degree C temperature stabilization scenarios and 2) the current population exposure of 414 cities around the world. For some locations with a modest projected increase in the height of an ESL event, the corresponding change in local population exposure is substantial. This suggests that physical ESL metrics may be poor surrogates for capturing some societal impacts. Overall, we find that impacts are highly localized and depend on the gradient of the population versus elevation profile over the range of elevations between the current and future ESL height. While population exposure is just one measure, considering a variety of human system, natural resource, and ecosystem-based outcomes may provide a more complete snapshot of coastal flood risk under different climate scenarios. Such an approach would improve upon existing methods used by the Intergovernmental Panel on Climate Change (IPCC).