Background: Response to proton pump inhibitor (PPI) treatment in dyspepsia is unpredictable. Aim: To identify symptoms associated with response to esomeprazole in order to target patients for empirical treatment. Methods: Eight hundred and five uninvestigated, primary care patients with upper GI symptoms that were considered to be acid-related were randomised to 2 weeks' treatment with esomeprazole 40 mg or placebo. The study population was divided into a model sample (N = 484) and a validation sample (N = 321). We developed a therapeutic index to predict PPI response from the model sample and tested this in the validation sample. Results: Response to PPI was found in 68% of patients (44% in placebo arm). Bothersome heartburn and early satiety were associated with increased likelihood of PPI response, whereas dull abdominal pain, pain relieved by bowel movements and nausea in women were associated with a decreased likelihood of PPI response. Patients in the validation sample could be classified as having a 'very high' (n = 55), 'high' (n = 123), 'medium' (n = 78) or 'low' (n = 65) probability of PPI response. The therapeutic gains over placebo were 55%, 31%, 20% and 22%, respectively. Conclusions In patients with uninvestigated dyspepsia, PPI responders can be reliably identified by a simple pocket chart using symptoms and patient characteristics (ClinicalTrials.gov NCT00318968).