© 2020, The Author(s). Purpose: The CRASH-3 trial is a randomised trial of tranexamic acid (TXA) on death and disability in patients with traumatic brain injury (TBI). It is based on the hypothesis that early TXA treatment can prevent deaths from post-traumatic intracranial bleeding. The results showed that timely TXA treatment reduces head injury deaths in patients with reactive pupils and those with a mild to moderate GCS at baseline. We examined routinely collected CT scans in a sample of 1767 CRASH-3 trial patients to explore if, why, and how patients are affected by TXA. Methods: The CRASH-3 IBMS is an explanatory study nested within the CRASH-3 trial. We measured the volume of intracranial bleeding on CT scans using established methods (e.g. ABC/2). Results: Patients with any un-reactive pupil had a median intracranial bleeding volume of 60 ml (IQR 18–101 ml) and patients with reactive pupils had a median volume of 26 ml (IQR 1–55 ml). Patients with severe GCS had median intracranial bleeding volume of 37 ml (IQR 3–75 ml) and patients with moderate to mild GCS had a median volume of 26 ml (IQR 0.4–50 ml). For every hour increase from injury to the baseline scan, the risk of new bleeding on a further scan decreased by 12% (adjusted RR = 0.88 [95% CI 0.80–0.96], p = 0.0047). Conclusion: Patients with reactive pupils and/or mild to moderate GCS may have benefited from TXA in the CRASH-3 trial because they had less intracranial bleeding at baseline. However, because bleeding occurs soon after injury, treatment delay reduces the benefit of TXA.