BACKGROUND: Poor geographical access to health services and routes to a cancer diagnosis such as emergency presentations have previously been associated with worse cancer outcomes. However, the extent to which access to GPs determines the route that patients take to obtain a cancer diagnosis is unknown. METHODS: We used a linked dataset of cancer registry and hospital records of patients with a cancer diagnosis between 2006 and 2010 across eight different cancer sites. Primary outcomes were defined as 'desirable routes to diagnosis' [screen-detected and 2-week wait (TWW) referrals] and 'less desirable routes' [emergency presentations and death certificate only (DCO)]. All other routes (GP referral, inpatient elective and other outpatient) were specified as the reference category. Geographical access was measured as travel time in minutes from patients to their GP, and multinomial logistic regression was used to estimate relative risk ratios (RRR). RESULTS: Longer travel was associated with increased risk of diagnosis via emergency and DCO, but decreased risk of diagnosis via screening and TWW. Patients travelling over 30 minutes had the highest risk of a DCO diagnosis, which was statistically significant for breast, colorectal, lung, prostate, stomach and ovarian cancers (compared with patients with travel times ≤10 minutes: RRR 5.89, 7.02, 2.30, 4.75, 10.41; P < 0.01 and 3.51, P < 0.05). DISCUSSION: Poor access to GPs may discourage early engagement with health services, decreasing the likelihood of screening uptake and increasing the likelihood of emergency presentations. Extra effort is needed to promote early diagnosis in more distant patients.