Abstract The International Reference Ionosphere (IRI) model can partially reproduce ionospheric long‐term trends (LTTs), although not intentionally, provided that an appropriate effective ionospheric index is used. We evaluate how model predictions driven by a solar activity proxy (R12) and by an effective ionospheric proxy (IG12) can reproduce the long‐term variability of the F2‐layer critical frequency (foF2) compared to hourly manually‐scaled data from the Rome ionosonde station (41.9°N, 12.5°E) over the period 1980–2022. After removing solar‐cycle and short‐term variability through empirical mode decomposition, a residual monotonic decreasing trend was identified for both measured and modeled foF2. While the R12‐driven model performs well until the late 1990s, it fails thereafter, whereas the IG12‐driven model reproduces LTTs over the entire period. The discrepancy is attributed to a post–23rd solar cycle change in the solar proxy–ionospheric response relationship, highlighting the need for effective indices and explicit time dependence in IRI.