jats:pThe present research applied a mixed-fidelity approach to examine the fan–intake interaction. Flow separation induced by a distortion generator (DG) is either resolved using large eddy simulation (LES) or modeled using the standard k–ω model, Spalart–Allmaras (SA) model, etc. The immersed boundary method with smeared geometry (immersed boundary method with smeared geometry (IBMSG)) is employed to represent the effect of the fan and a wide range of test cases is studied by varying the (a) height of the DG and (b) proximity of the fan to the DG. Comparisons are drawn between the LES and the Reynolds-averaged Navier–Stokes (RANS) approaches with/without the fan effect. It is found that in the “absence of fan,” the discrepancies between RANS and LES are significant within the separation and reattachment region due to the well-known limitations of the standard RANS models. “With the fan installed,” the deviation between RANS and LES decreases substantially. It becomes minimal when the fan is closest to the DG. It implies that with an installed fan, the inaccuracies of the turbulence model are mitigated by the strong flow acceleration at the casing due to the fan. More precisely, the mass flow redistribution due to the fan has a dominant primary effect on the final predictions and the effect of turbulence model becomes secondary, thereby suggesting that high fidelity eddy resolving simulations provide marginal improvements to the accuracy for the installed cases, particularly for the short intake–fan strategies with fan getting closer to intake lip.</jats:p>