It is commonly known that a drastic decrease in global carbon dioxide (CO2) emissions is necessary, in order to reach the climate goals set by the Paris agreement in 2015. A key technology towards achieving that goal is CCUS - Carbon, Capture, Utilisation, and Sequestration. By using supercritical CO2 instead of brine/water as a geothermal working fluid, geothermal energy production can possibly be expanded to regions with lower heat gradients in subsurface formations, while permanently storing CO2 underground. This first-order, conceptual study investigates the suitability of the Aquistore CCS-site for a CO2-circulation pilot test. For doing so, numerical simulations were performed to learn about the site responses to CO2-circulation, the amount of back-produced CO2 versus brine, and to estimate the flow behaviour in a potential CO2 gas production well. A key requirement for a successful CO2-circulation pilot test is to prevent liquid loading in the CO2 gas production well. Liquid loading occurs if brine or water accumulates in the production well. It can be avoided by maintaining an annular flow regime in the multi-phase fluid production stream of the production well. The resulting flow regime is mainly controlled by the total mass flow rate of the production stream. This in turn strongly depends on the overall transmissivity of the reservoir. The obtained simulation results suggest that steady-state conditions will occur within days to a few weeks after the start of the CO2-circulation. Moreover, our results show that the amount of back-produced CO2 is one order of magnitude larger than the amount of back-produced brine. In the majority of cases, we observe that the back-produced fluid production stream will ultimately flow in an annular flow pattern. Further analysis of CO2-circulation results indicate a need to better characterize the subsurface multiphase fluid flow behaviour. To this end, attempts to constrain the uncertainty associated with the Aquistore reservoir characterization and CO2 plume growth through high-resolution history matching of non-isothermal injection data and time-lapse seismic monitoring surveys are discussed.