Mitigating global climate change will require gigaton-scale carbon dioxide removal (CDR) as a supplement to rapid emissions reduction. The oceans cover 71% of the Earth surface and have the potential to provide much of the required CDR. However, none of the proposed marine CDR (mCDR) methods is sufficiently well understood to determine their real-world efficiency and environmental side effects. Here, we argue that using natural mCDR analogs should become the third interconnecting pillar in the mCDR assessment as they bridge the gap between numerical simulations (i.e., large scale/reduced complexity) and experimental studies (i.e., small scale/high complexity). Natural mCDR analogs occur at no cost, can provide a wealth of data to inform mCDR, and do not require legal permission or social license for their study. We propose four simple criteria to identify particularly useful analogs: 1) large scale, 2) abruptness of perturbation, 3) availability of unperturbed control sites, and 4) reoccurrence. Based on these criteria, we highlight four examples: 1) equatorial upwelling as a natural analog for artificial upwelling, 2) downstream of Kerguelen Island for ocean iron fertilization, 3) the Black and Caspian Seas for ocean alkalinity enhancement, and 4) the Great Atlantic Sargassum Belt for ocean afforestation. These natural analogs provide a reality check for experimental assessments and numerical modeling of mCDR. Ultimately, projections of mCDR efficacy and sustainability supported by observations from natural analogs will provide the real-world context for the public debate and will facilitate political decisions on mCDR implementation. We anticipate that a rigorous investigation of natural analogs will fast-forward the urgently needed assessment of mCDR.