Assessments of ecosystem functioning are a fundamental ecological challenge and an essential foundation for ecosystem-based management. Species trophic position (TP) is essential to characterize food web architecture. However, despite the intuitive nature of the concept, empirically estimating TP is a challenging task due to the complexity of trophic interaction networks. Various methods are proposed to assess TPs, including using different sources of organic matter at the base of the food web (the 'baseline'). However, it is often not clear which methodological approach and which baseline choices are the most reliable. Using an ecosystem-wide assessment of a tropical reef (Marquesas Islands, with available data for 70 coral reef invertebrate and fish species), we tested whether different commonly used TP estimation methods yield similar results and, if not, whether it is possible to identify the most reliable method. We found significant differences in TP estimates of up to 1.7 TPs for the same species, depending on the method and the baseline used. When using bulk stable isotope data, the choice of the baseline significantly impacted TP values. Indeed, while nitrogen stable isotope (δ15N) values of macroalgae led to consistent TP estimates, those using phytoplankton generated unrealistically low TP estimates. The use of a conventional enrichment factor (i.e. 3.4‰) or a 'variable' enrichment factor (i.e. according to feeding guilds) also produced clear discrepancies between TP estimates. TPs obtained with δ15N values of source amino acids (compound-specific isotope analysis) were close to those assessed with macroalgae. An opposite seasonal pattern was found, with significantly lower TPs in winter than in summer for most species, with particularly pronounced differences for lower TP species. We use the observed differences to discuss possible drivers of the diverging TP estimates and the potential ecological implications.