Ambient-noise seismic interferometry (ANSI) applied to passive body-wave measurements retrieves an estimate of the reflection response as if from a source at a receiver position. Often, the limited compliance with theoretical assumptions causes erroneous absolute amplitudes of the retrieved physical reflections, and additional artefacts. Nevertheless, the retrieved reflection data may be further used for time-lapse interpretation, since the latter exploits relative amplitude differences. Here, we study the feasibility of applying ANSI to time-lapse passive seismic data to extract the time-lapse reflection signal produced by the exploitation of a reservoir. We base our study on the case of the demonstration site for CO2 storage at Ketzin, Germany. With numerical experiments, we apply ANSI to two passive datasets using a base and a repeat scenario (after velocity decreases in the CO2 reservoir) and modelled by random distributions of band-limited noise sources. We show that the retrieval of an unambiguous time-lapse signal is enabled by increased common illumination between the two datasets. Finally, we apply ANSI by auto-correlation to Ketzin field data and show the retrieval of responses consistent with modelled and active field data. We conclude that ANSI applied to field data has the potential for time-lapse differences extraction and interpretation.