Creep and acoustic emission tests were conducted under uniaxial compression on sandstone samples with no (i.e. intact), one and two prefabricated fractures collected from the Xiaolangdi Reservoir area using a RLJW-2000 rock microcomputer-controlled rheology servotest system and PCI-2 acoustic emission testing equipment, respectively. The change laws of the acoustic emission amplitudes, peak frequencies, acoustic emission events and b values of samples with different fracture numbers (i.e. number of fractures) were systematically analysed. The differences between the prediction made by acoustic emission b values and cumulative ringing count for the precursor time of creep failure in a rock mass were studied comparatively, thereby identifying the effects of fracture number on the creep and acoustic emission characteristics of sandstone. According to the study results: (1) With an increase in the prefabricated fracture number, the acoustic emission signal counts of samples increased within the amplitude intervals of 60–80 and 80–100 dB. The signal counts were especially high in the high-amplitude interval of 80–100 dB but they decreased within the low-amplitude interval of 45–60 dB. (2) Under each stress level, the total acoustic emission event counts generated by the single- and double-fracture samples were 1.90 and 2.36 times that generated by the intact sample, respectively. Under the last stress level, the event counts generated by the single- and double-fracture samples were 4.64 and 6.19 times the event count generated by the intact sample, respectively. A higher prefabricated fracture number implied higher total acoustic emission event counts generated by samples and higher acoustic emission event counts generated under the last stress level. (3) Under the same stress level, the frequency band concentration of acoustic emission signals of samples became more pronounced as the prefabricated fracture number increased. Under the last stress level, as the prefabricated fracture number increased, the peak frequencies of the acoustic emission signals of samples were concentrated in a greater number of frequency ranges (four, seven and eight frequency ranges, respectively). (4) With an increase in the prefabricated fracture number, the average acoustic emission b values of samples presented an overall declining trend, with an increased Δ b range. (5) The plunge rates of acoustic emission b values of the single- and double-fracture samples before accelerated creep failure increased by 8.0 and 11.2%, respectively, compared with the intact sample. (6) The acoustic emission b value curves and cumulative ringing count curves for samples with different fracture numbers predicted creep failure in the rock mass earlier than the strain curves. For all samples, the precursor time points of creep failure in the rock mass predicted by acoustic emission b value curves were all earlier than those predicted by corresponding cumulative ringing count curves. As the prefabricated fracture number increased, the precursor time points of creep failure in the rock mass predicted by acoustic emission b value curves occurred even earlier. Thematic collection: This article is part of the Engineering Geology and Hydrogeology of the Anthropocene collection available at: https://www.lyellcollection.org/topic/collections/engineering-geology-and-hydrogeology-of-the-anthropocene