Photocatalytic overall water splitting represents a promising strategy for clean hydrogen production. However, its catalytic performance is limited by the back reaction, in which the produced H2 undesirably reacts with O2 to regenerate H2O. Although microscale techniques such as Cr2O3 overcoating on H2 evolution electrocatalysts have been the mainstream approach for suppressing the back reaction by blocking O2, their inherent limitations in toxicity, chemical instability, and versatility necessitate the development of alternative strategies. Herein, we propose a macrostructural engineering approach using a gas-diffusion-type reactor, in which inert Ar gas directly reached the Rh2O3/CoOOH/Al-doped SrTiO3 photocatalyst immobilized on a porous substrate. The photocatalytic performance was investigated under various conditions, including Ar flow rates and O2 mixing ratios, and compared with that of conventional suspension- and sheet-type reactors employing indirect gas flow. An apparent quantum yield of the gas-diffusion-type reactor achieved 25% at wavelength of 370 nm under atmospheric pressure, which was an order of magnitude higher than those of the conventional reactors (<2%) and comparable to that obtained with Cr2O3 coating over Rh2O3 (27%). Operando X-ray absorption spectroscopy under photocatalytic conditions and electrochemical measurements using an electrodeposited Rh2O3 model electrode clarified that the outstanding performance originated from maintaining a low local O2 concentration near the Rh2O3 sites via efficient removal of evolved gases by the direct Ar blowing. As a result, this study established the conceptual validation of gas-diffusion engineering for efficient photocatalytic overall water splitting. These findings provide a versatile and practical strategy for overcoming the limitations of back reactions in photocatalytic overall water splitting and potentially in other photoreactions.