The spatial pattern of geotagged tweets reflects the dynamic distribution of the ambient population during day and night as a result of people’s routine activities. A few studies have assessed the impact of tweets-derived ambient population on crime and the spillover effect of such impact at different spatial and temporal scales. However, none has revealed the intraday variation of such spillover effect. This study analyzes both the direct and spillover effects of tweets-derived ambient population on crime and its intraday difference in day and night during weekdays and weekends. Four crime types, including assault, burglary, robbery, and theft, are examined at the neighborhood level. The analysis is based on negative binomial regression models, with the control of necessary socioeconomic and land-use variables driven by criminology theories. Results show (1) tweets-derived ambient population affects the magnitude of crime, but this effect varies by types of crime at different time periods of the day and week, and (2) the spillover effect of the tweets-derived ambient population exists for all four types of crime during most of the time periods at the neighborhood level and is particularly pronounced for thefts at all time periods. Similar results are seen in the block-level analysis. This study further confirms the utility of the count of geotagged tweets as a measure of the ambient population and its spatial lag for intraday analyses of crime, particularly theft.