Localised aluminium contamination can lead to high concentrations in coastal waters, which have the potential for adverse effects on aquatic organisms. This research investigated the toxicity of 72-h exposures of aluminium to three marine diatoms (Ceratoneis closterium (formerly Nitzschia closterium), Minutocellus polymorphus and Phaeodactylum tricornutum) by measuring population growth rate inhibition and cell membrane damage (SYTOX Green) as endpoints. Toxicity was correlated to the time-averaged concentrations of different aluminium size-fractions, operationally defined as <0.025 μm filtered, <0.45 μm filtered (dissolved) and unfiltered (total) present in solution over the 72-h bioassay. The chronic population growth rate inhibition after aluminium exposure varied between diatom species. C. closterium was the most sensitive species (10% inhibition of growth rate (72-h IC10) of 80 (55–100) μg Al/L (95% confidence limits)) while M. polymorphus (540 (460–600) μg Al/L) and P. tricornutum (2100 (2000–2200) μg Al/L) were less sensitive (based on measured total aluminium). Dissolved aluminium was the primary contributor to toxicity in C. closterium, while a combination of dissolved and precipitated aluminium forms contributed to toxicity in M. polymorphus. In contrast, aluminium toxicity to the most tolerant diatom P. tricornutum was due predominantly to precipitated aluminium. Preliminary investigations revealed the sensitivity of C. closterium and M. polymorphus to aluminium was influenced by initial cell density with aluminium toxicity significantly (p < 0.05) increasing with initial cell density from 103 to 105 cells/mL. No effects on plasma membrane permeability were observed for any of the three diatoms suggesting that mechanisms of aluminium toxicity to diatoms do not involve compromising the plasma membrane. These results indicate that marine diatoms have a broad range in sensitivity to aluminium with toxic mechanisms related to both dissolved and precipitated aluminium.