Simple Summary One fundamental feature that has emerged from in vitro application of cold plasmas in cancer treatment is the key role of the liquid phase covering the cells. In the present work, we investigated the effect of direct and indirect plasma treatments on two cancer and three normal cell lines to assess the benefits of one treatment over the other in terms of death of tumor versus healthy cells. Our results demonstrate that indirect plasma treatment is as efficient at killing tumor cells as an appropriate combination of H2O2, NO2− and acidic pH in ad hoc solutions, while sparing normal cells. However, direct plasma treatment is far more efficient at killing normal than tumor cells, and we provide evidence that short- and long-lived reactive species contribute synergistically to kill normal cells, while having an additive effect regarding tumor cell death. Collectively, our results call the use of plasma-activated liquid in cancer treatment into question. Abstract (1) Plasma-activated liquids (PAL) have been extensively studied for their anti-cancer properties. Two treatment modalities can be applied to the cells, direct and indirect plasma treatments, which differ by the environment to which the cells are exposed. For direct plasma treatment, the cells covered by a liquid are present during the plasma treatment time (phase I, plasma ON) and the incubation time (phase II, plasma OFF), while for indirect plasma treatment, phase I is cell-free and cells are only exposed to PAL during phase II. The scope of this work was to study these two treatment modalities to bring new insights into the potential use of PAL for cancer treatment. (2) We used two models of head and neck cancer cells, CAL27 and FaDu, and three models of normal cells (1Br3, NHK, and RPE-hTERT). PBS was used as the liquid of interest, and the concentration of plasma-induced H2O2, NO2− and NO3−, as well as pH change, were measured. Cells were exposed to direct plasma treatment, indirect plasma treatment or reconstituted buffer (PBS adjusted with plasma-induced concentrations of H2O2, NO2−, NO3− and pH). Metabolic cell activity, cell viability, lipid peroxidation, intracellular ROS production and caspase 3/7 induction were quantified. (3) If we showed that direct plasma treatment is slightly more efficient than indirect plasma treatment and reconstituted buffer at inducing lipid peroxidation, intracellular increase of ROS and cancer cell death in tumor cells, our data also revealed that reconstituted buffer is equivalent to indirect plasma treatment. In contrast, normal cells are quite insensitive to these two last treatment modalities. However, they are extremely sensitive to direct plasma treatment. Indeed, we found that phase I and phase II act in synergy to trigger cell death in normal cells and are additive concerning tumor cell death. Our data also highlight the presence in plasma-treated PBS of yet unidentified short-lived reactive species that contribute to cell death. (4) In this study, we provide strong evidence that, in vitro, the concentration of RONS (H2O2, NO2− and NO3−) in combination with the acidic pH are the main drivers of plasma-induced PBS toxicity in tumor cells but not in normal cells, which makes ad hoc reconstituted solutions powerful anti-tumor treatments. In marked contrast, direct plasma treatment is deleterious for normal cells in vitro and should be avoided. Based on our results, we discuss the limitations to the use of PAL for cancer treatments.