Although freshly formed or unaltered biogenic aggregates are easily recognized, identifying the origin of aggregates altered by physical and biological processes remains empirical and prone to error. The aim of this study was to distinguish between biogenic (BIO) and physicogenic (PHYS) aggregates in various states of fragmentation or size classes using visual, physical and chemical characteristics. Casts produced by Amynthas khami (BIO) and surrounding soil aggregates without visible biological activity (PHYS) were left to disaggregate by natural rainfall events and then separated into five size classes of >10, 10–5, 5–2, 2–0.5 and <0.5 mm. We then analysed aggregate morphology, elemental and stable isotope composition and soil stability, and used near-infrared spectroscopy (NIRS) to determine their chemical characteristics. Although visual assessment is the method most commonly used in the field to distinguish between BIO and PHYS, our study found that the results obtained were always prone to error and that the classification was arbitrary for BIO and PHYS aggregates smaller than 5 and 2 mm in size, respectively. Soil structural stability was only useful for identifying BIO aggregates larger than 2 mm. While C content and δ13C in BIO were always different from PHYS, regardless of soil aggregate size, N content and δ15N were similar. NIRS was the most effective method because it clearly discriminated soil aggregates on the basis of size and origin. The NIRS characteristics of BIO were also more uniform than those of PHYS, suggesting that BIO aggregates have a simpler organization and as a consequence more homogeneous ecological functions. Thus, our findings suggest that information may be lost when only the physical aspect of aggregates is used to quantify the activity of ecosystem engineers in soil. After fragmentation, BIO aggregates become hidden and although it may be impossible to distinguish them visually from PHYS aggregates they retain some of their specific chemical characteristics.