Assessing fluid leakage rates through cracks in concrete structures is essential in many fields (nuclear safety, fluid storage, durability of structures). In previous works, the authors proposed experimental setups and procedures to study water and air leakage through cracked concrete specimens under Brazilian loading conditions. In the present contribution, the previous setup for air leakage is adapted to measure fluid pressure inside a so-called “realistic” crack representative of an actual crack surface geometry. The specimen is made by precisely casting a concrete crack surface. The procedure allows to realize it in two blocks (upper and lower) separable along this reproduced surface. The lower block is drilled with four thin cylindrical holes opening into the crack and which, once the specimen is installed on the press, are connected to pressure sensors. The specimen is inserted into an airflow circuit, and particular attention is paid to avoiding parasite leaks. Different inlet/outlet pressure conditions combined with different crack openings are imposed. Two different crack surface geometries are tested: plane and realistic. The results show non-linear air pressure distributions inside the realistic crack and non-negligible head losses at the inlet and outlet sections of the flow. Consequently, additional non-linear effects appear in the relationship between mass flow rate and squared pressure gradient due to the geometry of the realistic crack. These results demonstrate that measuring only the pressures in the upstream and downstream reservoirs, as classically done in standard protocols from the literature, can lead to a misinterpretation of the experimental results.