Residential and office buildings have the largest share in energy consumption, followed by transport and industry. At the same time, many buildings do not leverage all feasible opportunities to increase their energy efficiency. Particularly, the solutions influencing the behaviour of the end-users are lacking. In this paper, we present a novel semantics-empowered approach for motivating end-users towards the adoption of energy efficient lifestyles, based on recommendations provided through personalised applications and serious games. As a foundation of our approach, we have designed two semantic models to represent energy consumption and behavioural characteristics of consumers. The Energy Efficiency Semantic Model represents energy consumption data collected from a heterogeneous sensor network, while the Behavioural Semantic Model focuses on energy consumption profile of end-users. These models are being validated in the reference architecture and use cases of EU H2020 project ENTROPY.