There is still disagreement about whether to routinely use spectrophotometry to detect xanthochromia in cerebrospinal fluid (CSF) or whether visual inspection is adequate. We aimed to evaluate the diagnostic accuracy of these methods in detecting an aneurysmal subarachnoid hemorrhage in patients with sudden onset severe headache. When a patient presents to the emergency department with a headache for which there is suspicion of a subarachnoid hemorrhage, the gold standard to rule this out is to perform a CSF analysis for xanthochromia with or without spectrophotometry if the cranial non-contrast computed tomography (CT) upon admission is negative. Having applied the gold standard, we retrospectively included patients with acute headache who underwent both CT scan and CSF spectrophotometry at our hospital in the period 2002-2020. Patients were excluded if the cranial CT was interpreted as positive, there was a bloody CSF, or if visual assessment data of the CSF was unavailable. We scrutinized the patients' medical records and evaluated the benefit of spectrophotometry compared to visual inspection. The net bilirubin absorbance cut-off for support of subarachnoid hemorrhage was set at >0.007 absorbance units. The spectrophotometry was also considered positive if the net bilirubin absorbance was ≤0.007 and net oxyhemoglobin absorbance was ≥0.1 absorbance units. We calculated and compared the sensitivity and specificity of CSF spectrophotometry and visual inspection of the CSF. In total, 769 patients, with a mean age of 42.3 ± (standard deviation [SD] = 17.3) years, were included. The headache onset was classified as a thunderclap headache in 41.5%, and 4.7% had a sudden loss of consciousness. Fifteen patients (2%) were finally diagnosed with a subarachnoid hemorrhage, six (0.8%) had an aneurysmal subarachnoid hemorrhage, seven (0.9%) had a perimesencephalic hemorrhage, one (0.1%) had a cortical cerebral sinus venous thrombosis, and one (0.1%) had a spinal epidural hematoma. Four patients (0.5%) had a subarachnoid hemorrhage that was not detected by visual inspection, and two were caused by an aneurysmal rupture. One of these two patients died just before intervention, and the other underwent coiling for an anterior communicating aneurysm. The number needed for lumbar puncture to detect a subarachnoid hemorrhage was 51, but 128 to detect an aneurysmal hemorrhage. The corresponding numbers needed for CSF spectrophotometric analysis were 192 and 385, respectively. Spectrophotometry was positive in 31 patients (4.0%), of whom 18 (2.3%) also had visually detected xanthochromia (11 true positive). The mean net bilirubin absorbance in the 13 samples with visually clear CSF was 0.0111 ± (SD = 0.0103) absorbance units, compared to 0.0017 ± (SD = 0.0013) in the CSF with negative spectrophotometry. The corresponding figures for net oxyhemoglobin absorbance were 0.0391 ± (SD = 0.0522) versus 0.0057 ± (SD = 0.0081). The sensitivity of spectrophotometric xanthochromia detection was 100% (95% confidence interval [CI], 78-100), compared to 73% (95% CI, 45-92) for visual xanthochromia detection. The specificity of spectrophotometric xanthochromia detection was 98% (95% CI, 97-99) compared to 99% (95% CI, 98-100) for visual xanthochromia detection. Both methods had high negative predictive values: 100% (95% CI, 99.5-100) versus 99.5% (95% CI, 98.6-99.9), respectively. Both visual inspection and spectrophotometry have high diagnostic accuracy for detecting CSF xanthochromia, but the lower sensitivity of visual assessment makes it unreliable, and we recommend the use of spectrophotometry in clinical practice.