Most systematic projects to monitor bird populations, like breeding bird surveys, require large and coordinated volunteer networks that are lacking in many parts of the world such as the Global South. Data from less systematic citizen science (CS) programmes offer an alternative to data from systematic initiatives in these regions, but the semi-structured nature of such data also presents several challenges. The utility of semi-structured CS data to monitor bird species abundance is contingent on how, where, and how comparably birdwatchers watch birds, year on year. Trends inferred directly from the data can be confounded during years when birdwatchers may behave differently, such as during the COVID-19 pandemic. We wanted to ascertain how the data uploaded from India to one such CS platform, eBird, was impacted by this deadliest global pandemic of the 21st century. To understand whether eBird data from the pandemic years in India is useful and comparable to data from adjacent years, we explored several quantitative and qualitative aspects of the data (such as birdwatcher behaviour) at multiple spatial and temporal scales. We found no negative impact of the pandemic on data generation. Data characteristics changed largely only during the peak pandemic months characterised by high fatality rates and strict lockdowns, possibly due to decreased human mobility and social interaction. It remained similar to the adjacent years during the rest of this restrictive period, thereby reducing the impact of the aberrant peak months on any annual inference. Moreover, impacts on data characteristics varied widely across states in India, resulting in no strong consistent trend at the national level—unlike results from elsewhere in the world. Our findings show that birdwatchers in India as contributors to CS were resilient to disturbance, and that the effects of the pandemic on birdwatching effort and birdwatcher behaviour are highly scale- and context-dependent. In summary, eBird data in India from the pandemic years remains useful and interpretable for most large-scale applications, such as abundance trend estimation, but will benefit from preliminary data quality checks when utilised at a fine scale.