Climate change is likely to have altered the ecological functioning of past ecosystems, and is likely to alter functioning in the future; however, the magnitude and direction of such changes are difficult to predict. Here we use a deep-time case study to evaluate the impact of a well-constrained CO2-induced global warming event on the ecological functioning of dominant plant communities. We use leaf mass per area (LMA), a widely used trait in modern plant ecology, to infer the palaeoecological strategy of fossil plant taxa. We show that palaeo-LMA can be inferred from fossil leaf cuticles based on a tight relationship between LMA and cuticle thickness observed among extant gymnosperms. Application of this new palaeo-LMA proxy to fossil gymnosperms from East Greenland reveals significant shifts in the dominant ecological strategies of vegetation found across the Triassic–Jurassic transition. Late Triassic forests, dominated by low-LMA taxa with inferred high transpiration rates and short leaf lifespans, were replaced in the Early Jurassic by forests dominated by high-LMA taxa that were likely to have slower metabolic rates. We suggest that extreme CO2-induced global warming selected for taxa with high LMA associated with a stress-tolerant strategy and that adaptive plasticity in leaf functional traits such as LMA contributed to post-warming ecological success.