<b><i>Introduction:</i></b> Traditional transcutaneous oxygen (tcPO₂) measurements are affected by measurement drift, limiting accuracy and usability. The new potentially drift-free oxygen fluorescence quenching technique has been combined in a single sensor with conventional transcutaneous carbon dioxide (tcPCO₂) monitoring. This study aimed to validate optical tcPO₂ and conventional tcPCO₂ against arterial blood gas samples in preterm neonates and determine measurement drift. <b><i>Methods:</i></b> In this prospective observational study, during regular care, transcutaneous measurements were paired to arterial blood gases from preterm neonates aged 24–31 weeks of gestational age (GA) with an arterial catheter. Samples were included based on stability criteria and stratified for sepsis status. Agreement was assessed using the Bland-Altman analysis. Measurement drift per hour was calculated. <b><i>Results:</i></b> Sixty-eight premature neonates were included {median (interquartile range [IQR]) GA of 26 4/7 [25 3/7–27 5/7] weeks}, resulting in 216 stable paired samples. Agreement of stable samples in neonates without sepsis (<i>n</i> = 38) and with suspected sepsis (<i>n</i> = 112) was acceptable for tcPO₂ and good for tcPCO₂. However, in stable samples of neonates with sepsis (<i>n</i> = 66), tcPO₂ agreement (bias and 95% limits of agreement) was −32.6 (−97.0 to 31.8) mm Hg and tcPCO₂ agreement was 4.2 (−10.5 to 18.9) mm Hg. The median (IQR) absolute drift values were 0.058 (0.0231–0.1013) mm Hg/h for tcPO₂ and 0.30 (0.11–0.64) mm Hg/h for tcPCO₂. <b><i>Conclusion:</i></b> The accuracy of optical tcPO₂ in premature neonates was acceptable without sepsis, while electrochemically measured tcPCO₂ remained accurate under all circumstances. Measurement drift was negligible for tcPO₂ and highly acceptable for tcPCO₂.