Background Accurate identification of acute stroke by Emergency Medical Dispatchers (EMD) is essential for timely and purposeful deployment of Emergency Medical Services (EMS), and a prerequisite for operating mobile stroke units. However, precision of EMD stroke recognition is currently modest. Aims We sought to identify targets for improving dispatcher stroke identification. Methods Dispatch codes and EMS patient records were cross-linked to investigate factors associated with an incorrect dispatch code in a prospective observational cohort of 625 patients with a final diagnosis of acute stroke or transient ischemic attack (TIA), transported to our stroke center as candidates for recanalization therapies. Call recordings were analyzed in a subgroup that received an incorrect low-priority dispatch code indicating a fall or unknown acute illness ( n = 46). Results Out of 625 acute stroke/TIA patients, 450 received a high-priority stroke dispatch code (sensitivity 72.0%; 95% CI, 68.5–75.5). Independent predictors of dispatcher missed acute stroke included a bystander caller (aOR, 3.72; 1.48–9.34), confusion (aOR, 2.62; 1.59–4.31), fall at onset (aOR, 1.86; 1.24–2.78), and older age (aOR [per year], 1.02; 1.01–1.04). Of the analyzed call recordings, 71.7% revealed targets for improvement, including failure to recognize a Face Arm Speech Time (FAST) test symptom (21/46 cases, 18 with speech disturbance), or failure to thoroughly evaluate symptoms (12/46 cases). Conclusions Based on our findings, efforts to improve dispatcher stroke identification should primarily focus on improving recognition of acute speech disturbance, and implementing screening of FAST-symptoms in emergency phone calls revealing a fall or confusion. Clinical Trial Registration URL: http://www.clinicaltrials.gov . Unique identifier: NCT02145663.