Freshwater marsh restoration can be a viable natural climate solution; however, the extent to which marsh soils bury and preserve organic carbon within policy‐relevant timescales remains highly uncertain. Here, we compare organic carbon masses and accumulation rates from an undrained reference marsh, a passively restored freshwater marsh (reflooded after 1954) and a chronosequence of actively restored freshwater marshes (<10 years in age) situated in Lake Erie watersheds in the Long Point Biosphere Reserve of Ontario, Canada. The reference site has sustained the highest rates of short‐term organic carbon accumulation (235 g C m−2 yr−1) over the last four decades and has the highest mass of soil organic carbon (122 tC/ha) at 0–30 cm depth. Organic carbon masses are highly variable among all restored wetlands (16–115 tC/ha) at 0–30 cm depth and are not strongly related to time since restoration at least over the last 10 years. Nonetheless, we show that passive wetland restoration generates high rates of organic carbon accumulation (144 g m−2 yr−1) on a multidecadal scale where sites are low‐lying, underlain by alluvial deposits and connected to larger ground and surface water networks. Active restoration measures (e.g. excavation, installation of berms) may promote organic carbon preservation, particularly where fine‐grained soil texture promotes waterlogging. We demonstrate the importance of substrate, topographic gradient, and hydrology in selecting sites for marsh restoration to maximize carbon sequestration, and argue that the presettlement context and reference paleorecords provide necessary baselines for directing successful wetland restoration.