Purpose <p>Projecting greenhouse gas emissions into the future using prospective life cycle assessment (pLCA) requires a clear understanding of how the carbon intensity (CI) of product systems evolves over time. Yet, few studies have empirically examined how and why CI of product systems change over multi-decade periods. Using corn ethanol as a case-study, we assess how/why CI modelling estimates have changed over time in a longstanding model (GREET) to derive lessons for pLCA.</p> Methods <p>In our study, we conduct a 25-year retrospective analysis of US corn ethanol (1996–2022) by tracing the LCA model parameters from the GREET model. We reviewed more than 55 GREET reports, tracked changes in over 20 model input parameters, and quantified their individual contributions to CI changes. Changes fell into four emergent categories: (i) Energy and material efficiency, (ii) Knowledge of product system, (iii) Market and policy responses, and (iv) Background systems. Results were compared to historical projections for the CI of corn ethanol; the emergent categories were also assessed against a selection of pLCA studies (<i>N</i> = 46).</p> Results and discussion <p>The modelled CI of corn ethanol reduced 31% from 64 to 44&#xa0;g CO2eq/MJ over 25 years. Efficiency improvements (e.g., increases in corn and ethanol yields; energy use reductions in farming and biorefining), emerged as dominant drivers, accounting for 47% of the reduction. Updates to emission factors and methods (e.g., N₂O and lime emission factors) contributed 14%, while market and policy driven changes (e.g., wet-mill to dry-mill production) and background system shifts (e.g., electricity grid mix) accounted for 15% and 8%, respectively. Historical projections from GREET and U.S. EPA models predicted efficiency improvements with reasonable accuracy, but underpredicted key transitions such as the shift to dry-mill processing and the declining use of coal in ethanol production. Most reviewed pLCAs accounted for efficiency gains, but only 13% incorporated evolving environmental knowledge and even fewer accounted for market and policy responses.</p> Conclusions <p>Retrospective analysis of emissions is a powerful way to understand temporal impacts to inform pLCA studies. Existing pLCA work often centres on efficiency improvements in the focal technology (e.g., yields, energy intensity, capacity factors), but often underrepresents parallel improvements in incumbent/benchmark systems, updates to emission factors or methods, and market- and policy-mediated changes associated with scale-up. To improve methodological robustness, we recommend representing emergent change categories— Energy and material efficiency, Knowledge of product system, Market and policy responses, and Background systems — explicitly in pLCA frameworks.</p>