BACKGROUND: The converging biology between enveloped viruses and extracellular vesicles (EVs) has raised interest in the application of engineered EVs as antiviral therapeutics. Following the recent COVID-19 pandemic, EVs engineered with either the ACE2-receptor or Spike-protein have been proposed as strategy to either decoy SARS-CoV-2, or to compete with its cell entry. For generic use as a platform for future pandemic preparedness, a systematic and quantitative comparison of both strategies is required to assess their limitations and benefits across different variants of concern. METHODS: Here we generated EVs decorated with either the ACE2-receptor or the Spike-protein of (Wuhan)-SARS-CoV-2 and used single vesicle imaging for in-depth quantitative characterisation. These vesicles were then systematically tested for anti-viral activity across SARS-CoV-2 variants of concern using both, pseudotype and live virus cellular infection models including primary human bronchial and nasal explants. RESULTS: Spike-protein EVs or ACE2-EVs recovered from transiently transfected HEK293T cells comprised only a small fraction of the EV secretome (5% or 20%, respectively) and were primarily derived from the plasma membrane rather than multivesicular bodies. Redirecting intracellular trafficking of the Spike protein by mutating its transmembrane or subcellular localisation domains did not increase the yields of Spike-EVs. Both types of vesicles inhibited SARS-CoV-2 (D614G) in a dose dependent manner with kinetics and immunohistochemistry consistent with an inhibition at the initial cell entry stage. ACE2-EVs were more potent than Spike-EVs and at least 500-1000 times more potent than soluble antibodies in a pseudotype model. Surprisingly, ACE2-EVs switched from an inhibitory to an enhancer activity for the Omicron BA.1 variant whereas Spike-EVs retained their activity across all variants of concern. CONCLUSIONS: While our data show that both types of engineered EVs potently inhibit SARS-CoV, the decoy versus competition strategy may result in diverging outcomes when considering viral evolution into new variants of concern. While Spike-EVs retain their competition for receptor binding even against higher affinity viral Spike mutations, the formation of complexes between ACE2-EVs and the virus may not only result in inhibition by decoy. As EVs are actively internalised by cells themselves, they may shuttle the virus into cells, resulting in a productive alternative cell entry route for variants such as Omicron, that diverge from strict plasma membrane protease cleavage to the use of endosomal proteases for release of their genome.