This study's objectives were to describe community oncologists' beliefs about and confidence with geriatric care and to determine whether geriatric-relevant information influences cancer treatment decisions. Community oncologists were recruited to participate in 2 multisite geriatric oncology trials. Participants shared their beliefs about and confidence in caring for older adults. They were also asked to make a first-line chemotherapy recommendation (combination vs single-agent vs no chemotherapy) for a hypothetical vignette of an older patient with advanced pancreatic cancer. Each oncologist received one randomly chosen vignette that varied on 3 variables: age (72/84 years), impaired function (yes/no), and cognitive impairment (yes/no). Other patient characteristics were held constant. Logistic regression models were used to identify associations between oncologist/vignette-patient characteristics and treatment decisions. Oncologist response rate was 61% (n=305/498). Most oncologists agreed that "the care of older adults with cancer needs to be improved" (89%) and that "geriatrics training is essential" (72%). However, <25% were "very confident" in recognizing dementia or conducting a fall risk or functional assessment, and only 23% reported using the geriatric assessment in clinic. Each randomly varied patient characteristic was independently associated with the decision to treat: younger age (adjusted odds ratio [aOR], 5.01; 95% CI, 2.73-9.20), normal cognition (aOR, 5.42; 95% CI, 3.01-9.76), and being functionally intact (aOR, 3.85; 95% CI, 2.12-7.00). Accounting for all vignettes across all scenarios, 161 oncologists (52%) said they would offer chemotherapy. All variables were independently associated with prescribing single-agent over combination chemotherapy (older age: aOR, 3.22; 95% CI 1.43-7.25, impaired cognition: aOR, 3.13; 95% CI, 1.36-7.20, impaired function: aOR, 2.48; 95% CI, 1.12-5.72). Oncologists' characteristics were not associated with decisions about providing chemotherapy. Geriatric-relevant information, when available, strongly influences community oncologists' treatment decisions.