Climate change is continuously intensifying droughts. Extreme droughts are expected to reduce soil water content and thus, ecosystem functioning such as above-ground primary productivity. Nonetheless, results of experimental drought studies vary from no impact to a significant decrease in soil water content and/or productivity. We experimentally imposed extreme drought as 30 % and 50 % precipitation reductions using rainout shelters for four years in temperate grasslands and in the forest understory. We studied the concurrent impact of two intensities of extreme drought on the soil water content and above-ground primary productivity in the last experimental year (resistance). Furthermore, we observed resilience as the extent to which both variables differ from ambient conditions after the removal of the 50 % reduction. We show a systematic difference in response to extreme experimental drought between grasslands and the forest understory irrespective of the intensity of the extreme drought. Namely, extreme drought resulted in a significant decrease of the soil water content and productivity in grasslands but not in the forest understory. Interestingly, the negative impacts in the grasslands did not persist as evidenced by the fact that soil water content and productivity were similar to ambient conditions after the removal of the drought. Our results indicate that extreme drought on small spatial scales does not necessarily result in a concurrent soil water decrease in the forest understory, while this is the case for grasslands, with respective consequences for the resistance of productivity. Grasslands, however, can be resilient. Our study highlights that considering the response of the soil water content is key to understanding divergent productivity responses to extreme drought among different ecosystems.