This dissertation is composed of two related projects focused on assessing the metallicity distribution of circumgalactic gas at low redshift. The metallicity of the gas is a key property to characterize, since it is a direct diagnostic of the level of enrichment of the gas. This, in turn, informs us on the origin of the gas, e.g., metal-enriched outflows or inflows or metal-poor inflows. To determine the metallicity distribution of this gas, we assemble the largest samples to date of QSO absorbers known to probe circumgalactic medium (CGM) gas at z < 1; these absorbers are known as Lyman-limit systems (LLSs) and partial Lyman-limit systems (pLLSs). My dissertation work consists of two main surveys: (1) a blind survey of 55 CGM absorbers toward 140 QSOs (Chapters 2 and 3) and (2) a survey aimed to increase the sample of absorbers in the range 16.9 < log N(H I) < 19 (Chapters 4 and 5). The second survey was directly motivated by the results of the blind survey, which suggested a change in the metallicity distribution between the pLLSs and LLSs; however, the sample of LLSs in the blind survey was not large enough to robustly assess that conclusion.As part of this work, we develop a new method to robustly derive the metallicity of ionized gas using only low-resolution spectra of H I and Mg II (Chapter 2); this significantly reduces the observational cost of such studies by a factor ~10. In the second part of my thesis, we also apply a Bayesian MCMC approach to determine the metallicity, which in particular provides more robustly-derived errors on the metallicities.My dissertation shows unequivocally that metal-poor absorbers (gas with less than 10% or even 1–2% solar) are not rare in the z < 1 universe, contrary to what was previously thought. About 60% of the CGM absorbers (as probed by pLLSs and LLSs) are metal-poor. Galaxies are therefore host to large reservoirs of cool, dense, low-metallicity gas, even down to z < 1. We demonstrate a strong evolution of the metallicity with N(H I) between 16 and >21 dex, i.e., from very ionized to neutral gas. Across this range, the metallicity distribution changes from a unimodal distribution at log N(H I) > 19 (with an average around 25% solar metallicity) to a much broader and more complicated distribution at lower N(H I) (where we find a distribution that is at least bimodal with prominent peaks at 30% and 2–5% solar metallicity, and possibly another peak at <1% solar metallicity).These findings provide new constraints for cosmological simulations that we discuss in Chapter 5. We argue that the low-metallicity CGM absorbers have properties consistent with cold accretion flows from the IGM seen in cosmological simulations. The high-metallicity CGM absorbers likely probe outflows from the host galaxy, tidally-stripped gas from satellite galaxies, or recycling flows.