With many of the world’s poor engaged in agriculture, agricultural development programmes often aim to improve livelihoods through improved farming practices. Research on the impacts of agricultural technology interventions is dominated by comparisons of adopters and non-adopters. By contrast, in this literature study, we critically review how technology evaluation studies assess differentiated impacts in smallholder farming communities. We searched systematically for studies which present agricultural technology impacts disaggregated for poor and relatively better-off users (adopters). The major findings of our systematic review are as follows: (1) The number of studies that assessed impact differentiation was startlingly small: we were able to identify only 85, among which only 24 presented empirical findings. (2) These studies confirm an expected trend: absolute benefits are larger for the better-off, and large relative benefits among the poor are mostly due to meagre baseline performance. (3) Households are primarily considered as independent entities, rather than as connected with others directly or indirectly, via markets or common resource pools. (4) Explanations for impact differentiation are mainly sought in existing distributions of structural household characteristics. We collated the explanations provided in the selected studies across a nested hierarchy: the field, the farm or household, and households interacting at the farming system level. We also consider impact differentiation over time. With this, we provide a structured overview of potential drivers of differentiation, to guide future research for development towards explicitly recognizing the poor among the poor, acknowledging unequal impacts, aiming to avoid negative consequences, and mitigating them where they occur.