An experimental study was conducted to improve understanding of the interactions between bypassed oil and CO2, and to investigate their contribution to bypassed oil recovery. First, experiments were carried out that focus on the interaction in terms of light-hydrocarbon extraction and dissolution-induced oil swelling. These were studied with respect to the simulated shallow/deep bypassed oil, according to their distribution in the bypassed region. Core flooding tests were designed to determine the level of bypassed oil recovery, which focus on the interaction-dominated, soaking, and 'huff and puff' (HnP) strategies (traditional methods for recovering non-bypassed oil). We find that, significant amounts of shallow bypassed oil could be extracted with a maximum extraction efficiency of 85.2% at 40 MPa, making it shrink instead of the more commonly observed expansion. Such deep bypassed oil could not be extracted to the same significant extent as shallow bypassed oil, mainly behaving in a dissolution-dominated expansive manner. Core flooding tests show that interactions occurring during soaking only allow a small amount of bypassed oil to be recovered (yielding a somewhat disappointing recovery factor of approximately 20%). It is likely that mainly shallow bypassed oil is activated in this case. Additionally, the transition of the dominating interactions during soaking was also observed by oil composition analysis. Lastly, but fortunately, HnP is able to enhance bypassed oil recovery to a more satisfying level of around 90%, removing the majority of the deep bypassed oil therewith.