In this paper, the exergy performance of direct-expansion solar-assisted heat pump systems working with R22 and R433A (mixture of R290 and R1270, 70:30 by mass) was experimentally assessed. The experiments were carried out under the metrological conditions of Calicut in India. (Longitude and latitude of location are $$75.78 ^{\circ }\hbox {E}$$ and $$11.25^{\circ }\hbox {N}$$ , respectively.) The artificial neural network model was developed for simulating the performance of a direct-expansion solar-assisted heat pump system to have realistic performance comparison. The experimental data observed during the year 2016 were used for training and testing the performance of network. The results showed that the network predicted exergy performance of a direct-expansion solar-assisted heat pump was found to be closer to the experimental results with a maximum fraction of absolute variance, minimum root-mean-square values and coefficient of variance. The system exergy destruction of R22 and R433A was found to be 1.36 and 1.25 kW, respectively. Moreover, R433A is identified as an energy-efficient and environmental-friendly alternative to phase out R22 in solar-assisted heat pump systems.