Plant invasions drive biodiversity loss, transform ecosystems, and promote positive‐feedback cycles between invasion and fire. However, the long‐term impacts of invasive grasses across landscapes with diverse plant communities and interactions with fire are poorly known. Our objectives were to examine whether buffel grass (Cenchrus ciliaris), a globally significant plant invader, altered the abundance of understory and overstory plants, homogenized plant composition, and shifted ecosystems from woodlands to grassland and to explore interrelationships between invasion and fire. We combined two methodological approaches to assess invasion spread and impacts of buffel grass in the Aṉangu Pitjantjatjara Yankunytjatjara (APY) Lands of arid central Australia: a before‐after‐control‐impact (BACI) experiment over 25 years at 15 sites and a paired‐plot (randomized‐block) experiment at 18 sites. Both experiments spanned two geographic regions and multiple vegetation communities situated on flat plains and rocky hills. We used generalized linear mixed models to analyze predictions about plant abundance and permutational multivariate ANOVA (PERMANOVA) and permutational multivariate analysis of dispersion (PERMDISP) to examine changes in community composition. Fire and invasion interactions were explored using fire history or the relative fire tolerance of plant species as covariates, predictors, or responses. Fire interacted with the invasion process in multiple ways. Invaded sites had burnt more frequently and recently than native sites in one region, and where propagules were present in 1995, buffel grass abundance increased most when fires ensued. Abundance of understory plant functional groups (native grasses, ferns, and vines) decreased with invasion, and understory shrubs decreased due to frequent fires in invaded sites. Overstory composition shifted from fire‐sensitive species toward fire‐tolerant species, but this was not directly attributable to invasion. Partial evidence for ecosystem regime shifts included homogenization of understory communities in invaded rocky hills, and an increase in woody shrub cover at native but not invaded sites over 25 years, resulting in a 5% cover difference by 2019. Impacts were detected across heterogeneous ecological communities at a scale not previously tested amongst high background community variability. Although invasion is not dependent on fire, the acceleration of invasion spread and impacts with fire is a critical consideration for future research and management of grass invaders.