Designed artificial reefs (ARs) are deployed for various purposes including the enhancement of recreational fisheries. The ability to assess recreational harvest is important for determining the effectiveness of AR deployments. Harvest estimation at AR fisheries pose many logistical and budgetary challenges. We present a pragmatic approach to estimate harvest at an AR off coastal Sydney, Australia, that combines existing datasets and a cost-effective sampling design from two different time periods. Fishing effort data collected from June 2013 to May 2014 were derived directly from digital images of the AR and were validated by direct observation. Multiple datasets were then integrated to obtain a list of taxa that are harvested by recreational fishers within the AR area. Data from a series of probability-based surveys conducted prior to the deployment of the AR from March 2007 to February 2009 were used to obtain estimates of harvest rates for these taxa. Harvest at the reef was estimated by multiplying fishing effort and these harvest rates together. Total annual recreational harvest from the AR during June 2013–May 2014 was estimated to be 1016 ± 82 fish by number, 700 ± 59 kg of fish by weight, and 12,504 kg per km2. Standardized harvest at the Sydney AR was relatively high (2.3–43.6 times larger) compared to other fishery areas from which the fishable area is known. Harvest at the AR was dominated by 6 functional groups (ambush predators, leatherjackets, large to medium pelagic fish, small pelagic fish, medium demersal predators and large demersal predators), which accounted for 92% of the total annual harvest by number, and 95% of the total annual harvest by weight. Comparisons of standardized harvest between the Sydney AR and other fishery areas revealed two distinct groups, a) the AR and Swansea channel, a marine-dominated entrance to a large estuary, and b) all other fishery areas. The use of existing datasets from a previous time period to represent current conditions in a fishery can be subject to potential bias since harvest composition and harvest rates were calculated using data collected prior to the implementation of the AR. However, this pragmatic approach may be the only viable option when the implementation of probability-based survey methods is logistically complex and prohibitively costly. Future studies attempting to estimate harvest at small, discrete AR fisheries located near large population centers should therefore consider an integrated methodology that combines existing datasets and cost-effective sampling designs.