Crop genetic diversity underpins the adaptive capacity of food systems but is declining. Maintaining diversity is challenging because most crop varieties underperform the best-suited ones under current conditions, therefore the private value of currently low-yielding varieties to farmers is limited. However they may carry traits that become critical as conditions change, so their public value as a genetic resource is large and long-term. Payments for Agrobiodiversity Conservation Services (PACS) – adapted from Payments for Ecosystem Services – aims to address this disparity by compensating farmers’ opportunity cost of conserving declining landraces. Here we provide the first counterfactual evaluation of a PACS scheme, using competitive bidding to allocate community conservation contracts among 461 farmers. We focus on enset (Ensete ventricosum), a staple food for over 20 million Ethiopians, which is almost entirely absent from ex situ collections. Of 141 enset landraces, 70% were threatened and 26% locally extinct. Using two complementary difference-in-differences designs we show that, relative to controls, PACS increased richness by 4.9 landraces per household, doubling typical farm diversity, and increased the number of households cultivating contracted landraces tenfold. Rather than crowding out non-contracted landraces, PACS improved their conservation status, with larger gains in communities with a greater share of contracted landraces. This indicates motivational crowding in, with PACS delivering conservation gains beyond those paid for. Scaling average compensation (US$8.46 per household) suggests that securing threatened enset diversity nationally would require US$141,000-640,000, <0.1% of Ethiopia’s agriculture and food security budget. PACS offers a cost-effective, scalable strategy for in situ agrobiodiversity conservation.