Summary Background In patients with a first eradication failure, a second (rescue) therapy still fails in > 20% of cases. Aim To compare rifabutin and levofloxacin rescue regimens in patients with two consecutive Helicobacter pylori eradication failures. Methods Patients, in whom first treatment with omeprazole–clarithromycin–amoxicillin and a second trial with omeprazole–bismuth–tetracycline–metronidazole (or ranitidine bismuth citrate with these antibiotics) had failed, received 10 days of treatment with either rifabutin (150 mg b.d.) or levofloxacin (500 mg b.d.), plus amoxicillin (1 g b.d.) and omeprazole (20 mg b.d.). Cure rates were evaluated by the 13C‐urea breath test. Results Twenty patients received rifabutin, and 20 levofloxacin. All the patients returned for follow‐up. Compliance in the rifabutin group was 100%. Four patients in the levofloxacin group did not take the medication correctly (in two cases due to adverse effects: myalgia and rash). Side effects in the rifabutin and levofloxacin groups were reported in 60% and 50% of the cases, respectively. Five patients (25%) treated with rifabutin presented with leucopenia, and six (30%) treated with levofloxacin presented with myalgias. Per‐protocol cure rates were 45% (95% confidence interval, 26–66%) in the rifabutin group, and 81% (57–93%) in the levofloxacin group (P < 0.05). Intention‐to‐treat cure rates were, 45% (26–66%) and 85% (64–95%), respectively (P < 0.01). Conclusions After two previous H. pylori eradication failures, a 10‐day triple levofloxacin‐based rescue regimen is more effective than the same regimen with rifabutin.