Even one metre seems generous. A reanalysis of data in: Chu et al. (2020) Physical distancing, face masks, and eye protection to prevent person-to-person transmission of SARS-CoV-2 and COVID-19.

Abstract

Re-examination of the large dataset collected and meta-analysed by Dr Chu and his colleagues contradicts their conclusions about the effects of separation distance on infection risk. Their conclusion was based on misunderstandings of the datasets. Each of these estimated risk relative to that incurred when touching infected individuals. Allowing for this suggests that the main advantage of social distancing, a perhaps 78% (95% CI 24, 92) reduction in risk of infection, occurs at distances below 1m. The data imply an 11% chance of further distances reducing the risk, with any effects likely to be small. However the limitations of the dataset do limit the strength of these conclusions.