The CO2 concentration is a good indicator of the indoor air quality, and the ground-level CO2 concentration needs to be measured precisely in order to monitor global warming. Several designs for a cost-effective CO2 sensor in order to replace the commercialized infrared spectroscopic CO2 sensor have been suggested. Among them, the potentiometric-type sensor that consists of a Na+ (or Li+) conductor, a metal carbonate auxiliary phase and a reference electrode is promising because it shows a satisfactory EMF (electromotive force) over a wide CO2 concentration range. In order to commercialize the electrochemical CO2 sensor, minimizing the signal drift is crucial. The EMF of a potentiometric sensor is the result of the Na+ activity difference between a CO2-sensitive sensing electrode (SE) and a CO2-insensitive counter electrode (CE). As the sensor ages in air, Na+-ions migrate spontaneously from the SE toward the CE, which forms various Na-containing deposits by a reaction with the humidity, O2 and CO2. This makes it difficult to define the Na+ activity at the CE, which is a major reason for the signal drift. There have been two different approaches to fix the Na+ activity at the CE: One is the placement of the CE in the reference gas such as CO2-free air, which requires tight gas sealing between the SE and CE. The other is the attachment of a two-phase mixture to the CE, which fixes the Na+-ion activity thermodynamically.