By using a multimethod strategy, we seek to generate more rigorous evidence on landscape restoration and its impacts at the household level. Using pre-restoration remote-sensing data, a machine-learning algorithm is used for the identification of similar pieces of land to Action Against Desertification (AAD) restoration sites. Comparison households were then selected from communities bordering these sites through a replication of the AAD targeting process. Finally, the impact analysis is based on propensity score adjustment techniques, applied to survey data. Overall findings suggest that participation in landscape restoration influenced household-livelihood strategies towards climate-resilient options, including a reduction of crop sales accompanied by an increase in the commercialization of livestock and livestock by-products. Households also planted more trees on their individual land, because of the restoration of communal and public lands. While this occurred without harming food security, we don’t observe a substantial increase in food security within treatment households. This suggests that food security support could be strengthened as part of restoration activities and/or that impacts of opportunity-led diversification may need a longer period to accrue. Larger impacts observed within the early takers of the programme reinforce these conclusions. Overall, the analysis also provides an innovative approach to ex-post evaluation settings.