We present a new, detailed analysis of late-time mid-infrared observations of the Type II-P supernova (SN) 2003gd. At about 16 months after the explosion, the mid-IR flux is consistent with emission from 4 x 10^(-5) M☉ of newly condensed dust in the ejecta. At 22 months emission from pointlike sources close to the SN position was detected at 8 and 24 μm. By 42 months the 24 μm flux had faded. Considerations of luminosity and source size rule out the ejecta of SN 2003gd as the main origin of the emission at 22 months. A possible alternative explanation for the emission at this later epoch is an IR echo from preexisting circumstellar or interstellar dust. We conclude that, contrary to the claim of Sugerman and coworkers, the mid-IR emission from SN 2003gd does not support the presence of 0.02 M☉ of newly formed dust in the ejecta. There is, as yet, no direct evidence that core-collapse supernovae are major dust factories.