The atmospheric characterization of exoplanets through transit spectroscopy has become one of the standard techniques for exploring the interiors of these distant worlds in the past few years. It requires exquisite spectrophotometric precisions most instruments were never designed for, which has in turn pushed them to the limits of their capabilities in order to overcome instrumental systematics that hamper the detections of exoplanet atmospheric signatures. The Hubble Space Telescope (HST) has been one of the main workhorses for transit spectroscopy studies to date, providing the best window to explore these small signals. Despite enormous progress on understanding the instruments onboard it, however, no general model for the systematic trends found on transit light curves obtained with its instruments has been found yet, which limits the precision of some of the most exciting datasets taken with it. Thus, the challenge of precisely accounting and modeling those systematics to achieve higher precision in current atmospheric detections—and improve the noise floor for future ones—must be solved in order to fully exploit the capabilities of this and future instruments such as the upcoming James Webb Space Telescope (JWST). Here we report our efforts to employ the technique of Pixel-Level Decorrelation (PLD) to observations of exoplanets performed with HST. In PLD, the data itself is used to remove the systematics, without using external models or parameters. This technique has been tested for photometry but not yet for spectrophotometry. We attempted to use this spectroscopic PLD—sPLD—with HST observations of exoplanet HD 189733b taken with both the STIS and WFC3 instruments, and were successful in decorrelating STIS data with sPLD to similar precision to one of the more complex models found in the literature. We believe the technique shows great promise not only for HST/STIS observations in general, but for future observations with JWST, which will show similar 2D spectroscopic structure to that of HST/STIS.