This technical note examines K2-18b as a diagnostic case in exoplanet atmospheric characterization. The aim is not to resolve the physical nature of the planet, nor to assess biosignatures as such, but to distinguish between molecular claims that remain comparatively persistent across recent analyses and claims whose status changes substantially with data reduction, wavelength binning, model-space definition, and retrieval choices. Recent work on K2-18b makes this distinction unusually visible. Across the current literature, CH4 emerges as the most comparatively persistent molecular inference. CO2 appears less stable, being supported in some analyses of the near-infrared and combined datasets and weakened or lost in others. DMS and/or DMDS remain more fragile still, especially in the mid-infrared, where unresolved systematics, binning dependence, and reference-model sensitivity strongly affect interpretation. The resulting point is diagnostic and deliberately modest: not every preferred molecule is already a stable molecular claim. In this sense, K2-18b serves as a useful benchmark not only for atmospheric interpretation, but also for assessing the differential stability of retrieved claims under changing observational and analytical conditions.
Danilo Tavella (2026) studied this question.