Self-report personality inventories may be useful in directing perpetrators of intimate partner violence (IPV) to appropriate intervention programs. They may also have predictive capabilities in assessing the likelihood of desistance or persistence of IPV. However, validity problems are inherent in self-report clinical tools, particularly in forensic settings. Scores of the modifying indices (subsections of the scale designed to detect biases in responding) of the Millon Clinical Multiaxial Inventory-III (MCMI-III) often are not reported in research. This study analyses the response sets of a sample of 492 IPV perpetrators at intake to a Danish perpetrator program. Profiles were grouped into levels of severity, and the proportion of exaggerated or minimized profiles at each severity level was analyzed. Findings suggested that 30% of the present sample were severely disturbed or exaggerating their symptoms. As expected, there were significant levels of exaggerated profiles present in the severe pathology group and significant levels of minimized profiles in the low pathology group. Self-referred participants were more likely to exaggerate their pathology, but minimization was not associated with referral status. Nor was there an association between gender and the modifying indices. It is suggested that so-called “fake good” or “fake bad” profiles should not necessarily be treated as invalid, but that elevations in the modifying indices can be interpreted as clinically and forensically relevant information in their own right and should be reported on in research.