<p>Hyperoxemia is common in acute brain injury, but its clinical significance is unclear. Most supporting evidence comes from retrospective studies that reduce oxygen exposure to static metrics — mean, peak, or time-above-threshold PaO₂ — obscuring the fact that patients with identical averages can have very diff erent exposure patterns: stable versus oscillating between hyperoxemic peaks and normoxic troughs. These patterns may not be biologically equivalent. Intermittent hyperoxia may activate adaptive Nrf2 or HIF-1α pathways resembling ischemic preconditioning, while sustained hyperoxemia may overwhelm them. Consistent with this, protocolized intermittent hyperoxia (hyperbaric or time-limited normobaric) has improved outcomes in severe traumatic brain injury and intracerebral hemorrhage in randomized trials, whereas uncontrolled sustained hyperoxemia is linked to harm in observational cohorts — particularly in subarachnoid hemorrhage, ischemic stroke, and post-cardiac-arrest patients, with no consistent harm signal in traumatic brain injury. We argue that future trials should move beyond liberal-versus-conservative comparisons toward metrics capturing the rate, frequency, and duration of oxygen fluctuations. Until then, sustained extreme hyperoxemia (PaO₂ 300 mmHg) should be avoided, while protocolized, time-limited hyperoxic exposure merits further study in patients with metabolic distress.</p>