Abstract Global compilations of zircon geochemistry have been used as evidence for changes in plate tectonic styles and surface environments. In particular, zircon δ18O has been used as a proxy for global sediment subduction and incorporation into igneous melts. However, research employing such compilations commonly ignores geologic and geographic context. We analyze a newly georeferenced zircon δ18O database from 1000 Ma to present. The compilation shows positive and negative isotopic excursions, both of which have been interpreted in the context of global phenomena. Sensitivity testing demonstrates that these excursions are the product of regional tectonic environments. Specifically, low Tonian–Cryogenian δ18O, previously interpreted as a climate signal of snowball Earth glacial meltwater, is isolated to ultra-high-pressure rocks of the Dabie Sulu orogen in central Asia, ophiolites in Egypt, and juvenile arc development in Madagascar, predating Cryogenian glaciation. Positive anomalies in the Ediacaran–Cambrian and Devonian, previously interpreted to record snowball Earth erosion and the rise of land plants, are the result of upper-plate, supracrustal sources of Gondwana. A Neogene anomaly is the result of volcanic rocks associated with Yellowstone and Iceland. Sampling bias in zircon geochemical compilations is analogous to that of Lagerstätten (sites of exquisite fossil preservation) and the “monograph effect” (large contributions from individual researchers) in paleontology, which both result in artifacts of over-representation. Long-term zircon δ18O trends broadly track the generation of collisional granites and continental arcs through time, recording the melting of sediments and sedimentary rocks in individual orogens, and not variations in sediment subduction globally.