Abstract Gastropods are commonly preserved as steinkerns (internal casts), a mode of fossilization that leads to loss of external morphological features. This loss of information is problematic for taxonomic identification and ecological inference in evaluating assemblages where original shell material is not preserved. We seek to quantify how closely gastropod steinkerns represent the morphology of their original shells. We investigated this relationship experimentally by fabricating steinkerns in silicone from modern gastropod shells and comparing their geometry to that of the shells we used to create them. In addition to recording traces of ornamentation such as ribs and spines, we used a theoretical morphospace framework to evaluate the fidelity of shell-coiling parameters in steinkerns. Our results show that some morphotypes reflect their taxonomic identification more accurately than others, indicating that steinkern fidelity is highly variable. Experimental steinkerns consistently cluster less reliably by morphotype than their original shell counterparts. Additionally, we find that shell thickness is an important factor in determining steinkern fidelity. The fidelity of the high-spired Duplicaria duplicata, for example, is significantly lower than the average value for the morphotypes investigated whereas the fidelity of planispiral Haplotrema concavum and open-coiling Epitonium is significantly higher, a trend related to shell thickness. Thus, taxonomic identification and subsequent analyses, such as community composition, of steinkern assemblages must recognize this differential fidelity to counter preservational biases.