Changes in the risk of exposure to infectious disease agents can be tracked through variations in antibody prevalence in vertebrate host populations. However, information on the temporal dynamics of the immune status of individuals is critical. If antibody levels persist a long time after exposure to an infectious agent, they could enable the efficient detection of the past circulation of the agent; if they persist only a short time, they could provide snap shots of recent exposure of sampled hosts. Here, we explored the temporal dynamics of seropositivity against Lyme disease agent Borrelia burgdorferi sensu lato ( Bbsl ) in individuals of a widespread medium‐sized mammal species, the roe deer ( Capreolus capreolus ), in France. Using a modified commercially available immunoassay we tested 1554 blood samples obtained in two wild deer populations monitored from 2010 to 2020. Using multi‐event capture‐mark‐recapture models, we estimated yearly population‐, age‐, and sex‐specific rates of seroconversion and seroreversion after accounting for imperfect detection. The yearly seroconversion rates indicated a higher level of exposure in early (2010–2013) than in late years (2014–2019) to infected tick bites in both populations, without any detectable influence of sex or age. The relatively high rates of seroreversion indicated a short‐term persistence of antibody levels against Bbsl in roe deer. This was confirmed by the analysis of samples collected on a set of captive individuals that were resampled several times a few weeks apart. Our findings show the potential usefulness of deer as a sentinel for tracking the risk of exposure to Lyme disease Bbsl , although further investigation on the details of the antibody response to Bbsl in this incompetent host would be useful. Our study also highlights the value of combining long‐term capture‐mark‐recapture sampling and short‐time analyses of serological data for wildlife populations exposed to infectious agents of relevance to wildlife epidemiology and human health.