ABSTRACT This study of police emergency calls in the UK addresses the interactional work conducted when dealing with reports of kidnap. In the UK, kidnap is classed as a type of ‘crime-in-action’, known to be complex to categorise and code for the appropriate police response. Using the qualitative method of ‘conversation analysis’, we address this complexity through analysing a dataset of anonymised emergency calls which are, at some point during the call or subsequent police investigation, categorised as ‘kidnap’. Analysing the calls, their categorisations and the accompanying incident logs, we aim to understand the difficulties that can arise in identifying this type of high-stakes incident at the first point of police contact. We find callers encode differing levels of ‘entitlement’ in requests for police assistance, with potential effects on call-handlers’ decisions about kidnap categorisations. We also observe interactional difficulties in establishing information about the incident, either through the caller’s displayed lack of knowledge or certainty, difficulty in producing turns or sometimes resistance to providing further information. These features may render the call-handler’s task of categorising incidents as ‘kidnap’ more challenging. Our identification of these communicative patterns has potential benefits for call-handlers’ practices in the police control room, providing an evidence-base from real-life talk for training. The findings also have implications at an institutional level, as they shed light on the negotiations that underly ‘categorisation’ work in policing, where there may sometimes only be a tacit understanding of how crime categories are decided during initial reports from the public.