• We use the multi-epoch, mid-infrared Spitzer Deep, Wide-Field Survey to investigate the variability of 474,179 objects in 8.1 deg^2 of the NDWFS Bootes field. We perform a Difference Image Analysis of the four available epochs between 2004 and 2008, focusing on the deeper 3.6 and 4.5 micron bands. We find that 1.1% of the studied sample meet our standard selection criteria for being classed as a variable source. We require that the 3.6 and 4.5 micron light-curves are strongly correlated (r>0.8) and that their joint variance exceeds that for all sources with the same magnitude by 2 sigma. We then examine the mid-IR colors of the variable sources and match them with X-ray sources from the XBootes survey, radio catalogs, 24 micron-selected AGN candidates, and spectroscopically identified AGNs from the AGN and Galaxy Evolution Survey (AGES). Based on their mid-IR colors, most of the variable sources are AGNs (76%), with smaller contributions from stars (11%), galaxies (6%), and unclassified objects. Most of the stellar, galaxy and unclassified sources are false positives. For our standard selection criteria, 11-12% of the mid-IR counterparts to X-ray sources, 24 micron-selected AGN candidates and spectroscopically identified AGNs show variability. Mid-IR AGN variability can be well described by a single power-law structure function with a power-law index of 0.5 at both 3.6 and 4.5 microns, and an amplitude of 0.1 mag on rest-frame time scales of 2 years. The variability amplitude is higher for shorter rest-frame wavelengths and lower luminosities. (Abridged)