Wind turbine development is growing exponentially and faster than other sources of renewable energy worldwide. While multi-turbine facilities have small physical footprint, they are not free from negative impacts on wildlife. This is particularly true for bats, whose population viability can be threatened by wind turbines through mortality events due to collisions. Wind turbine curtailment (hereafter referred to as “blanket curtailment”) in non-winter periods at low wind speeds and mild temperatures (i.e. when bats are active and wind energy production is low) can reduce fatalities, but show variable and incomplete effectiveness because other factors affect fatality risks including landscape features, rain, turbine functioning, and seasonality. The combined effects of these drivers, and their potential as criteria in algorithm-based curtailment, have so far received little attention. We compiled bat acoustic data recorded over four years at 34 wind turbine nacelles in France from post-construction regulatory studies, including 8,619 entire nights (251±58 nights per wind turbine on average). We modelled nightly bat activity in relation to its multiple drivers for three bat guilds, and assessed whether curtailment based on algorithm would be more efficient to limit bat exposure than blanket curtailment based on various combinations of unique wind speed and temperature thresholds. We found that landscape features, weather conditions, seasonality, and turbine functioning determine bat activity at nacelles. Algorithm-based curtailment is more efficient than blanket curtailment, and has the potential to drastically reduce bat exposure while sustaining the same energy production. Compared to blanket curtailment, the algorithm curtailment reduces average exposure by 20 to 29% and 7 to 12% for the high-risk guilds of long- and mid-range echolocators, and by 24 to 31% for the low-risk guild of short-range echolocators. These findings call for the use of algorithm curtailment as both power production and biodiversity benefits will be higher in most situations.