OBJECTIVE: To examine how much oncologists tell patients about the survival benefit of palliative chemotherapy during consultations at which decisions about treatment are made. DESIGN: Qualitative study in which consultations were observed and digitally recorded. SETTING: Teaching hospital and district general hospital in south west England. PARTICIPANTS: 37 patients with advanced non-small cell lung cancer (n=12), pancreatic cancer (n=13), and colorectal cancer (n=12); and nine oncologists, including four consultants and five registrars. MAIN OUTCOME MEASURES: All recordings were transcribed completely, anonymised, and electronically coded with ATLAS.ti. Constant comparison was used to identify themes and patterns. The framework method of data management, in which data were charted, was used to aid transparency of interpretation. RESULTS: During the consultations, information given to patients about survival benefit included numerical data ("about four weeks"), an idea of timescales ("a few months extra"), vague references ("buy you some time"), or no mention at all. In most consultations (26/37) discussion of survival benefit was vague or non-existent. CONCLUSIONS: Most patients were not given clear information about the survival gain of palliative chemotherapy. To aid decision making and informed consent, we recommend that oncologists sensitively describe the benefits and limitations of this treatment, including survival gain.