Background: The optimal approach for adult patients hospitalized with severe and critical coronavirus disease 2019 (COVID-19), non-responsive to antiviral and immunomodulatory drugs, is not well established. Our aim was to evaluate feasibility and safety of extracorporeal photopheresis (ECP) in this setting. Methods: A prospective, single-center investigational study was performed between 2021 and 2022 at a tertiary referral center for COVID-19. Patients diagnosed with COVID-19 were screened, and cases with severe or critical disease fulfilling pre-defined clinical and biochemical criteria of non-response for >5 days, despite remdesivir, dexamethasone and immunomodulation (tocilizumab, baricitinib, ruxolitinib), were consecutively enrolled. After patient inclusion, two ECP sessions on two consecutive days per week for 2 weeks were applied. Patients were followed-up per protocol from study inclusion, and clinical, virological and radiological outcomes were assessed at the end of treatment (EOT) +28 days. Results: A total of seven patients were enrolled. At inclusion, four out of seven (57.1%) were admitted to the ICU, all patients had ongoing cytokine storm. Additionally, 3/7 (42.9%) had radiological progression on chest CT. At EOT+28 days, 2/7 (28.6%) patients died due to non-ECP-related causes. Among the survivors, no additional requirement for intensive care unit admission or radiological progression was observed, and invasive mechanical ventilation could be weaned off in 1/5 (20.0%). All patients achieved whole-blood SARS-CoV-2 RNAemia clearance, while 3/7 (42.9%) no longer showed detectable respiratory SARS-CoV-2 RNA. According to immune biomarker profiling, ECP mainly facilitated a decrease in plasma IL-6 and IL-17A levels, as well as the physiological regeneration of peripheral blood immunocyte subpopulations, notably CD8+/CD45RO+ memory T-cells. No safety signals were identified. Conclusions: ECP appears to be a safe and feasible option for adults hospitalized with severe or critical COVID-19 who do not respond to pharmacological interventions. Further trial data are warranted to assess its optimal use. Trial registration: ClinicalTrials.gov NCT05882331 (retrospectively registered).