In the present work we investigate the thermal diffusion behavior of three different binary mixtures with a thermal lens (TL) setup. In the setup used in this study we avoid the addition of a dye for systems, such as aqueous mixtures, with a weak absorption band at a wavelength of 980 nm. In some aqueous systems with a complex phase behavior the addition of dye significantly affects the apparent measured thermal diffusion properties. The studied systems are dimethylsulfoxide (DMSO) in water, the ionic liquid 1-ethyl-3-methylimidazolium ethylsulfate (EMIES) in butanol and a non-ionic surfactant hexaethylene glycol monododecyl ether (C12E6) in water. The Soret coefficients of the selected systems cover a range of two orders of magnitude. For DMSO in water with a very low Soret coefficient of the order of ST∼ 10−3K−1 we find for a low DMSO content (c = 0.33) a reasonable agreement with previous measurements, while the weak thermal lens signal for the DMSO-rich mixture (c = 0.87) leads to 20% too large Soret coefficients with an uncertainty of more than 30%. Secondly we studied a liquid salt 1-ethyl-3-methylimidazolium ethylsulfate (EMIES) in butanol with a roughly ten times higher Soret coefficient of ST∼ 10−2K−1. For this system we performed additional measurements with another experimental technique, the classical thermal diffusion forced Rayleigh scattering (TDFRS), which requires the addition of a small amount of dye to increase the absorption. In the entire investigated concentration range the results obtained with the TL and classical TDFRS technique agree within the error bars. As a third system we studied a non-ionic surfactant hexaethylene glycol monododecyl ether (C12E6) in water with a Soret coefficient of the order of ST∼ 10−1K−1. For this system we find good agreement with previous measurements. We conclude that the TL technique is a reliable method for systems with a strong optical contrast and fairly large Soret coefficient of the order of ST∼ 10−2K−1.