ModaLens introduces a paired image‑swap audit to assess how the presence of radiology reports affects a vision‑language model’s reliance on the associated chest X‑ray image. Using MedGemma‑27B, the method evaluates 3,199 paired MIMIC‑CXR cases from 293 patients, each with 14 clinical questions, while swapping images but keeping the report and question constant. The results quantify changes in image sensitivity when reports are available versus when they are not.