Impairments in cerebral autoregulation (CA) are related to poor clinical outcome. Near infrared spectroscopy (NIRS) is a non-invasive technique applied to estimate CA. Our general purpose was to study the clinical feasibility of a previously published ‘NIRS-only’ CA methodology in a critically ill intensive care unit (ICU) population and determine its relationship with clinical outcome. Bilateral NIRS measurements were performed for 1–2 h. Data segments of ten-minutes were used to calculate transfer function analyses (TFA) CA estimates between high frequency oxyhemoglobin (oxyHb) and deoxyhemoglobin (deoxyHb) signals. The phase shift was corrected for serial time shifts. Criteria were defined to select TFA phase plot segments (segments) with ‘high-pass filter’ characteristics. In 54 patients, 490 out of 729 segments were automatically selected (67%). In 34 primary neurology patients the median (q1–q3) low frequency (LF) phase shift was higher in 19 survivors compared to 15 non-survivors (13° (6.3–35) versus 0.83° (−2.8–13), p = 0.0167). CA estimation using the NIRS-only methodology seems feasible in an ICU population using segment selection for more robust and consistent CA estimations. The ‘NIRS-only’ methodology needs further validation, but has the advantage of being non-invasive without the need for arterial blood pressure monitoring.