Objectives To assess whether workplace exposures as estimated via a COVID-19 Job Exposure Matrix (JEM) are associated with SARS-CoV-2. Methods Data on 244,470 participants were available from the ONS Coronavirus Infection Survey (CIS) and 16,801 participants from the Virus Watch Cohort, restricted to workers aged 20 to 64. Analysis used logistic regression models with SARS-CoV-2 as the dependent variable for eight individual JEM domains (number of workers, nature of contacts, contact via surfaces, indoor or outdoor location, ability to social distance, use of face covering, job insecurity, migrant workers) with adjustment for age, sex, ethnicity, Index of Multiple Deprivation (IMD), region, household size, urban vs rural area, and health conditions. Analyses were repeated for three time periods (i) February 2020 (Virus Watch)/April 2020 (CIS) to May 2021), (ii)June 2021 to November 2021, (iii) December 2021 to January 2022. Results Overall, higher risk classifications for the first six domains tended to be associated with an increased risk of infection, with little evidence of a relationship for domains relating to proportion of workers with job insecurity or migrant workers. By time there was a clear exposure-response relationship for these domains in the first period only. Results were largely consistent across the two cohorts. Conclusions An exposure-response relationship exists in the early phase of the COVID-19 pandemic for number of contacts, nature of contacts, contacts via surfaces, indoor or outdoor location, ability to social distance and use of face coverings. These associations appear to have diminished over time.