Stream restoration has increasingly been used as a best management practice for improving water quality in urbanizing watersheds, yet few data exist to assess restoration effectiveness. This study examined the longitudinal patterns in carbon and nitrogen concentrations and mass balance in two restored (Minebank Run and Spring Branch) and two unrestored (Powder Mill Run and Dead Run) stream networks in Baltimore, Maryland, USA. Longitudinal synoptic sampling showed that there was considerable reach-scale variability in biogeochemistry (e.g., total dissolved nitrogen (TDN), dissolved organic carbon (DOC), cations, pH, oxidation/reduction potential, dissolved oxygen, and temperature). TDN concentrations were typically higher than DOC in restored streams, but the opposite pattern was observed in unrestored streams. Mass balances in restored stream networks showed net uptake of TDN across subreaches (mean ± standard error net uptake rate of TDN across sampling dates for Minebank Run and Spring Branch was 420.3 ± 312.2 and 821.8 ± 570.3 mg m−2 d−1, respectively). There was net release of DOC in the restored streams (1344 ± 1063 and 1017 ± 944.5 mg m−2 d−1 for Minebank Run and Spring Branch, respectively). Conversely, degraded streams, Powder Mill Run and Dead Run showed mean net release of TDN across sampling dates (629.2 ± 167.5 and 327.1 ± 134.5 mg m−2 d−1, respectively) and net uptake of DOC (1642 ± 505.0 and 233.7 ± 125.1 mg m−2 d−1, respectively). There can be substantial C and N transformations in stream networks with hydrologically connected floodplain and pond features. Assessment of restoration effectiveness depends strongly on where monitoring is conducted along the stream network. Monitoring beyond the stream-reach scale is recommended for a complete perspective of evaluation of biogeochemical function in restored and degraded urban streams.