Human activities threaten reef ecosystems globally, forcing ecological change at rates and scales regarded as unprecedented in the Holocene. These changes are so profound that a cessation of reef accretion (reef ‘turn-off’) and net erosion of reef structures is argued by many as the ultimate and imminent trajectory. Here, we use a regional scale reef growth dataset, based on 76 core records (constrained by 211 radiometric dates) from 22 reefs along and across the inner-shelf of the Great Barrier Reef, Australia, to examine the timing of different phases of reef initiation (‘turn-on’), growth and ‘turn-off’ during the Holocene. This dataset delineates two temporally discrete episodes of reef-building over the last 8500 years: the first associated with the Holocene transgression-early highstand period [∼8.5–5.5 k calibrated years bp (cal ybp)]; the second since ∼2.3 k cal ybp. During both periods, reefs accreted rapidly to sea level before entering late evolutionary states – states naturally characterized by reduced coral cover and low accretion potential – and a clear hiatus occurs between these reef-building episodes for which no records of reef initiation exist. These transitions mimic those projected under current environmental disturbance regimes, but have been driven entirely by natural forcing factors. Our results demonstrate that, even through the late Holocene, reef health and growth has fluctuated through cycles independent of anthropogenic forcing. Consequently, degraded reef states cannot de facto be considered to automatically reflect increased anthropogenic stress. Indeed, in many cases degraded or nonaccreting reef communities may reflect past reef growth histories (as dictated by reef growth–sea level interactions) as much as contemporary environmental change. Recognizing when changes in reef condition reflect these natural ‘turn-on’– growth –‘turn-off’ cycles and how they interact with on-going human disturbance is critical for effective coral reef management and for understanding future reef ecological trajectories.