Octahedral metal complexes can bind to double strand (ds) DNA either by intercalation or by insertion, this latter mechanism being observed in the case of mismatched base pairs (bps). In this work we modeled the process of deintercalation from the major groove for Δ-Ru[(bpy)2(dppz)](2+) (1) and Δ-Rh[(bpy)2(chrysi)](3+) (2), prototypical examples of metallo-intercalators and metallo-insertors, respectively. By using advanced sampling techniques, we show that the two complexes have comparable deintercalation barriers and that in both systems the main cost of deintercalation is due to disruption of π-π stacking interactions between the intercalating moiety and the bps flanking the binding site. A striking difference between dppz and chrysi is found in their intercalation modes, being their longest axes, respectively, perpendicular and parallel to the P-P direction between opposite DNA strands. This leads the two ligands to deintercalate from the DNA through different mechanisms. Compound 1 goes through the formation of a metastable short-lived intermediate, with an overall free energy barrier of ~14.5 kcal/mol, in line with experimental findings. Due to the length of the dppz intercalating moiety, an extended plateau appears in the free energy landscape at ~3 kcal/mol above the most stable minimum. Compound 2 must cross a similar barrier (~15.5 kcal/mol), but does not form intermediates along the deintercalation path, and the deintercalation profile is steeper than that found for 1. Thus, the shape of the intercalating moiety affects the deintercalation mechanism of these inorganic molecules. This work is a first step to rationalize from a computational perspective the factors tuning the preferential binding mode of inorganic molecules (such as diagnostic probes, therapeutic agents, or regulators of DNA expression) to ds DNA.