We present a new method of discovering galaxy-scale, strongly-lensed QSO systems from unresolved light curves using the autocorrelation function. The method is tested on five rungs of simulated light curves from the Time Delay Challenge 1 that were designed to match the light-curve qualities from existing, ongoing, and forthcoming time-domain surveys such as the Medium Deep Survey of the Panoramic Survey Telescope And Rapid Response System 1, the Zwicky Transient Facility, and the Rubin Observatory Legacy Survey of Space and Time. Among simulated lens systems for which time delays can be successfully measured by current best algorithms, our method achieves an overall true positive rate of 28-58% for doublyimaged QSOs (doubles) and 36-60% for quadruply-imaged QSOs (quads) while maintains 10% false positive rates. We also apply the method to observed light curves of 22 known strongly-lensed QSOs, and recover 20% of doubles and 25% of quads. The tests demonstrate the capability of our method for discovering strongly-lensed QSOs from major time domain surveys. The performance of our method can be further improved by analysing multi-filter light curves and supplementing with morphological, colour, and/or astrometric constraints. More importantly, our method is particularly useful for discovering small-separation strongly-lensed QSOs, complementary to traditional imaging-based methods.