Carbon offsets allow greenhouse gas emitters to comply with an emissions cap by paying others outside of the capped sectors to reduce emissions. The first major carbon offset programme, the United Nations’ Clean Development Mechanism (CDM), has been criticized for generating a large number of credits from projects that do not actually reduce emissions. Following the controversial CDM experience, California pioneered a second-generation compliance offset programme that shifts the focus of quality control from assessments of individual projects to the development of offset protocols, which define project type-specific eligibility criteria and methods for estimating emissions reductions. We assess the ability of California’s ‘standardized approach’ to mitigate the risk of over-crediting greenhouse gas reductions by reviewing the development of two California offset protocols – Mine Methane Capture and Rice Cultivation. We examine the regulator’s treatment of three sources of over-crediting under the CDM: non-additional projects, inflated counterfactual baseline scenarios, and perverse incentives that inadvertently increase emissions. We find that the standardized approach offers the ability to reduce, but not eliminate, the risk of over-crediting. This requires careful protocol-scale analysis, conservative methods for estimating reductions, ongoing monitoring of programme outcomes, and restricting participation to project types with manageable levels of uncertainty in emission reductions. However, several of these elements are missing from California’s regime, and even best practices result in significant uncertainty in true emission reductions. Relying on carbon offsets to lower compliance costs risks lessening total emission reductions and increases uncertainty in whether an emissions target has been met.