Stablecoin transfers are often interpreted as payments. On programmable blockchains, however, they are frequently embedded in atomically executed transaction bundles that combine trading, lending, arbitrage, liquidity provision, and settlement. We show that ignoring this structure materially distorts the interpretation of stablecoin activity. Using 593 million event logs from 141 million Ethereum transactions involving three major U.S. dollar stablecoins, we develop a replicable framework to measure transaction complexity from archive node data, public contract labels, and event signatures. The analysis combines measures of token and contract co-usage, action type, computational complexity, urgency, and timing. Two results emerge. First, complexity is a first-order feature of stablecoin activity: nearly 60 percent of transfer events occur within complex transactions. Second, the three stable coins are not used interchangeably: their use differs systematically across transaction structures, urgency, and timing, consistent with distinct institutional designs and economic functions. Analyses that treat transfers as standalone payments therefore risk misclassifying a large share of on-chain stablecoin use, with implications for empirical measurement, market monitoring, and policy.