Abstract: CO2 huff-n-puff shows great potential to promote shale oil recovery after primary depletion. However, the extracting process of shale oil residing in different types of pores induced by the injected CO2 remains unclear. Moreover, how to saturate shale core samples with oil is still an experimental challenge, and needs a recommended procedure. These issues significantly impede probing CO2 huff-n-puff in extracting shale oil as a means of enhanced oil recovery (EOR) processes. In this paper, the oil saturation process of shale core samples and their CO2 extraction response with respect to pore types were investigated using online T1–T2 nuclear magnetic resonance (NMR) spectroscopy. The results indicated that the oil saturation of shale core samples rapidly increased in the first 16 days under the conditions of 60 °C and 30 MPa and then tended to plateau. The maximum oil saturation could reach 46.2% after a vacuum and pressurization duration of 20 days. After saturation, three distinct regions were identified on the T1–T2 NMR spectra of the shale core samples, corresponding to kerogen, organic pores (OPs), and inorganic pores (IPs), respectively. The oil trapped in IPs was the primary target for CO2 huff-n-puff in shale with a maximum cumulative oil recovery (COR) of 70% original oil in place (OOIP) after three cycles, while the oil trapped in OPs and kerogen presented challenges for extraction (COR < 24.2% OOIP in OPs and almost none for kerogen). CO2 preferentially extracted the accessible oil trapped in large IPs, while due to the tiny pores and strong affinity of oil-wet walls, the oil saturated in OPs mainly existed in an adsorbed state, leading to an insignificant COR. Furthermore, COR demonstrated a linear increasing tendency with soaking pressure, even when the pressure noticeably exceeded the minimum miscible pressure, implying that the formation of a miscible phase between CO2 and oil was not the primary drive for CO2 huff-n-puff in shale.