Summary Miscible injection of carbon dioxide has seen a significant increase in interest for the purpose of enhanced oil recovery (EOR) in conventional oil reservoirs. However, naturally fractured reservoirs, which are among the largest oil reserves in the world, are considered poor candidates for this process because of presumed low-performance efficiency. This paper presents the results of an experimental study that explains the effect of connate water saturation, matrix permeability and oil viscosity on the performance of gravity drainage from the matrix (into fracture) when it is surrounded by a CO2-filled fracture. Experiments were performed in an experimental model under different operating pressures to cover both immiscible and miscible conditions. Experiments were conducted using synthetic oil (nC10) and light crude oil in two Berea cores having large differences in permeability. In addition, the effect of connate water saturation was studied by performing experiments in an initially brine-saturated Berea core and comparing the results with those obtained when the core was 100% saturated with oil. The experimental results showed that matrix permeability had a significant effect on the rate of gravity drainage when CO2 was injected under immiscible conditions. When experiments were performed at immiscible conditions, production rate by gravity drainage was nearly five times greater in the Berea core with 1,000 md permeability compared to the core permeability of 100 md. The production rates in the cores investigated were similar at low pressures (below 3,400 kPa), but slightly higher for the higher-permeability core. As system pressure was increased beyond 3,400 kPa, the production rate from the higher-permeability core increased significantly, compared to the lower-permeability case. Beyond miscibility conditions (~6,900 kPa), matrix permeability was less significant, indicating the important role of capillary pressure in the gravity drainage mechanism. However, ultimate oil recovery was less sensitive to the matrix permeability at pressures near or above minimum miscibility pressure. The observations were more interesting when experiments were performed in the presence of connate water saturation. The ultimate oil recovery from a core saturated with oil in the presence of connate water saturation was less at immiscible conditions. However, at near-miscible and miscible conditions, the presence of connate water was beneficial to the gravity drainage mechanism in that it led to higher ultimate oil recovery. The effect of oil viscosity appeared to be important during the sustained miscibility of CO2 and hydrocarbon phases. For the crude oil examined, the heavier components that remain in the oil phase after the vapourizing gas drive limited the length of the oil production period when compared with the nC10 production. Miscible CO2 injection in fractured reservoirs is a viable option for both oil recovery and storage purposes because as the residual oil saturation is reduced, additional pore volume (PV) becomes available to store CO2 in its supercritical form. However, under immiscible conditions, when CO2 is injected at pressures below the minimum miscibility pressure (MMP) and above the supercritical condition, it is not beneficial for improving oil recovery by gravity drainage. This was clearly seen when gravity drainage experiments using crude oil were performed and MMP was not achieved at the maximum possible operating pressures. The results obtained from this study address the knowledge gap in the best practices for utilizing CO2 for improving oil recovery from fractured reservoir environments and demonstrate the effects of key parameters on the gravity drainage mechanism.