Antiviral agents blocking SARS-CoV-2 viral replication are desperately needed to complement vaccination to end the COVID-19 pandemic. Viral replication and assembly are entirely dependent on two viral cysteine proteases: 3C-like protease (3CLpro) and the papain-like protease (PLpro). PLpro also has deubiquitinase (DUB) activity, removing ubiquitin (Ub) and Ub-like modifications from host proteins, disrupting the host immune response. 3CLpro is inhibited by many known cysteine protease inhibitors, whereas PLpro is a relatively unusual cysteine protease, being resistant to blockade by such inhibitors. A high-throughput screen of biased and unbiased libraries gave a low hit rate, identifying only CPI-169 and the positive control, GRL0617, as inhibitors with good potency (IC50 < 10 lower case Greek μM). Analogues of both inhibitors were designed to develop structure-activity relationships; however, without a co-crystal structure of the CPI-169 series, we focused on GRL0617 as a starting point for structure-based drug design, obtaining several co-crystal structures to guide optimization. A series of novel 2-phenylthiophene-based non-covalent SARS-CoV-2 PLpro inhibitors were obtained, culminating in low nanomolar potency. The high potency and slow inhibitor off-rate were rationalized by newly identified ligand interactions with a 'BL2 groove' that is distal from the active site cysteine. Trapping of the conformationally flexible BL2 loop by these inhibitors blocks binding of viral and host protein substrates; however, until now it has not been demonstrated that this mechanism can induce potent and efficacious antiviral activity. In this study, we report that novel PLpro inhibitors have excellent antiviral efficacy and potency against infectious SARS-CoV-2 replication in cell cultures. Together, our data provide structural insights into the design of potent PLpro inhibitors and the first validation that non-covalent inhibitors of SARS-CoV-2 PLpro can block infection of human cells with low micromolar potency.