The balance between photosynthetic organic carbon production and respiration controls atmospheric composition and climate 1,2. The majority of organic carbon is respired back to carbon dioxide in the biosphere, but a small fraction escapes remineralization and is preserved over geologic timescales 3. By removing reduced carbon from Earth's surface, this sequestration process promotes atmospheric oxygen accumulation 2 and carbon dioxide removal 1. Two major mechanisms have been proposed to explain organic carbon preservation: selective preservation of biochemically unreactive compounds 4,5 and protection resulting from interactions with a mineral matrix 6,7. While both mechanisms can play a role across a range of environments and timescales, their global relative importance on 10 3-to 10 5-year timescales remains uncertain 4. Here we present a global dataset of the distributions of organic carbon activation energy and corresponding radiocarbon ages in soils, sediments, and dissolved organic carbon; we find that activation energy distributions broaden over time in all mineral-containing samples. This result requires increasing bondstrength diversity, consistent with the formation of organo-mineral bonds 8 but inconsistent with selective preservation. Radiocarbon ages further reveal that high-energy, mineralbound organic carbon persists for millennia relative to low-energy, unbound organic carbon. Our results provide globally coherent evidence for the proposed 7 importance of mineral protection in promoting organic carbon preservation. We suggest that similar studies of bond-strength diversity in ancient sediments may elucidate how and why organic carbon preservation-and thus atmospheric composition and climate-has varied over geologic time. Two classes of mechanisms-selectivity and protection-have been proposed to explain why some organic carbon (OC) escapes remineralization in soils and sediments 4-7. Biochemical selectivity hypotheses state that intrinsically bioavailable compounds such as sugars and amino acids are rapidly respired, whereas "recalcitrant" (macro)molecules such as lignin are selectively preserved due to their low energy yield, large size, and/or a lack of enzymes that can decompose them 4,5. Selective preservation has been extensively documented in dissolved OC (DOC) 9 , decaying woody tissue 10 , and sapropel sediments containing almost exclusively organic matter 5. In contrast, protection hypotheses state that particles shield OC from respiration regardless of intrinsic recalcitrance, potentially due to occlusion within pore spaces that are inaccessible to microbes and their extracellular enzymes 4,8,11-14. Specifically, protection often involves inspiration was always invaluable. We thank the National Ocean Sciences Accelerator Mass Spectrometer staff, especially A