We present ALMA Band 9 observations of the [ C II ] 158 \mu m emission for a sample of 10 main-sequence galaxies at redshift z \sim 2 , with typical stellar masses ( \log M _ { \star } /M _ { \odot } \sim 10.0–10.9 ) and star formation rates ( \sim 35–115 M _ { \odot } yr ^ { -1 } ) . Given the strong and well understood evolution of the interstellar medium from the present to z = 2 , we investigate the behaviour of the [ C II ] emission and empirically identify its primary driver . We detect [ C II ] from six galaxies ( four secure , two tentative ) and estimate ensemble averages including non detections . The [ C II ] -to-infrared luminosity ratio ( L _ { [ C~ { } { \small II } ] } / L _ { IR } ) of our sample is similar to that of local main-sequence galaxies ( \sim 2 \times 10 ^ { -3 } ) , and \sim 10 times higher than that of starbursts . The [ C II ] emission has an average spatial extent of 4 – 7 kpc , consistent with the optical size . Complementing our sample with literature data , we find that the [ C II ] luminosity correlates with galaxies ’ molecular gas mass , with a mean absolute deviation of 0.2 dex and without evident systematics : the [ C II ] -to-H _ { 2 } conversion factor ( \alpha _ { [ C~ { } { \small II } ] } \sim 30 M _ { \odot } /L _ { \odot } ) is largely independent of galaxies ’ depletion time , metallicity , and redshift . [ C II ] seems therefore a convenient tracer to estimate galaxies ’ molecular gas content regardless of their starburst or main-sequence nature , and extending to metal-poor galaxies at low- and high-redshifts . The dearth of [ C II ] emission reported for z > 6 –7 galaxies might suggest either a high star formation efficiency or a small fraction of UV light from star formation reprocessed by dust .