The prediction of reactivity is one of the long-standing objectives of chemistry. We have extracted reactivity patterns observed in aromatic molecules spanning 150 years of synthetic developments and used the data to test the predictive capacity of popular reactivity models. This systematic analysis has exposed numerous regioselectivities that are not predicted by resonance theory, electrostatic potentials or frontier molecular orbital theory. In contrast, calculated local ionisation energy surfaces are shown to consistently reveal the most nucleophilic sites in aromatic molecules even where established reactivity models fail. Furthermore, these local ionisation energy minima are found to correlate with experimentally determined reactivity parameters. Since ionisation energy surfaces are simple to interpret and are provided as standard in popular computational chemistry software, the approach serves as a readily accessible tool for visualising the fundamental factors governing the reactivity of aromatic molecules.