Summary 1. Globally, temperate grasslands have been heavily modified by agricultural intensification. The ecological integrity of many remaining semi‐natural grasslands is further threatened by the encroachment of invasive alien woody plant species, such as conifers. 2. Although the ecological impact of conifers is often inferred to be directly related to increasing conifer density at invaded sites, interpretation of these density‐dependent effects on native biodiversity is almost always confounded by strong correlations between density and time since invasion. Here, we isolate the direct effects of exotic conifer density on native invertebrate assemblages from the confounding effects due to time since invasion, by establishing replicated experimental plantings of the invasive Pinus nigra (densities of 400, 800, 1600, 2500 and 3500 trees per ha, established in 1993) in New Zealand. 3. From the 29 549 invertebrates captured, the relative abundance of major classes and orders was largely unaffected by conifer invasion at densities below 800 trees per ha at age 14 years, but differed substantially in higher‐density conifer stands (canopy cover >50%). At the species level, beetle species composition was highly sensitive to conifer invasion at densities as low as 400 trees per ha at age 14 years. Changes in beetle species composition were correlated with reduced soil moisture, increases in canopy cover, and trap distance from the nearest semi‐natural grassland. 4. Synthesis and applications. The effects of exotic conifer invasion on grassland invertebrate assemblages were strongly dependent on conifer density. In our study, the negative effects of conifer invasion on biodiversity were comparatively low at densities below ≤800 trees per ha at 14 years after invasion. To conserve invertebrate diversity in semi‐natural grasslands we recommend a dual focus of conservation actions targeting: (i) immediate control, or thinning, of high density conifer infestations (>800 trees per ha) to restore habitat to a state suitable for grassland species, and (ii) longer‐term removal of sparse, pre‐coning infestations in order to prevent further spread. We propose the use of canopy cover as a proxy for the confounded relationship between conifer density and time since invasion.