This paper investigates the environmental benefits of ride-sharing through its CO2 emission mitigation potential. Ride-sharing is expected to substantially decrease CO2 emissions by raising vehicle occupancy, thus mechanically reducing the number of vehicles on the road. Yet, as ride-sharing entails both a decrease in travel (monetary) costs and in travel times (inasmuch as it reduces road congestion), it is likely to make the car more attractive ultimately. This could result in mode switching in the short run (as travelers forsake public transport or active modes for car), as well as in longer distances travelled in the medium run. In the long run, people could even take advantage of the easier travel conditions to relocate further within the metropolitan area. To account for these rebound effects, we develop an integrated land-use transport model. This intends to capture the effects of ride-sharing on the whole household decision process regarding transport and residential location. The model is applied to the Paris region, with several ride-sharing scenarios for year 2030. While ride-sharing does indeed strongly reduce CO2 emissions, we find substantial rebound mechanisms. In contrast to the (naïve) expectation that raising vehicle occupancy by 50% would reduce CO2 emissions by 33%, the various rebound effects end up dividing the CO2 emission savings by a factor ranging from 2 to 3 depending on the day period considered (i.e. the morning or evening peak period). The rebound mechanisms - the mode switching, distance and relocation effects - should therefore be heeded. Some policy recommendations are provided to develop ride-sharing while curbing these unintended effects.