Abstract Aim To perform a direct, double‐blind, randomised, crossover comparison of subcutaneous and intravenous glucagon‐like peptide‐1 (GLP‐1) in hyperglycaemic subjects with type 2 diabetes naïve to GLP‐1‐based therapy. Materials and Methods Ten fasted, hyperglycaemic subjects (1 female, age 63 ± 10 years [mean ± SD], glycated haemoglobin 73.5 ± 22.0 mmol/mol [8.9% ± 2.0%], both mean ± SD) received subcutaneous GLP‐1 and intravenous saline, or intravenous GLP‐1 and subcutaneous saline. Infusion rates were doubled every 120 min (1.2, 2.4, 4.8 and 9.6 pmol·kg −1 ·min −1 for subcutaneous, and 0.3, 0.6, 1.2 and 2.4 pmol·kg −1 ·min −1 for intravenous). Plasma glucose, total and intact GLP‐1, insulin, C‐peptide, glucagon and gastrointestinal symptoms were evaluated over 8 h. The results are presented as mean ± SEM. Results Plasma glucose decreased more with intravenous (by ~8.0 mmol/L [144 mg/dL]) than subcutaneous GLP‐1 (by ~5.6 mmol/L [100 mg/dL]; p < 0.001). Plasma GLP‐1 increased dose‐dependently, but more with intravenous than subcutaneous for both total (∆ max 154.2 ± 3.9 pmol/L vs. 85.1 ± 3.8 pmol/L; p < 0.001), and intact GLP‐1 (∆ max 44.2 ± 2.2 pmol/L vs. 12.8 ± 2.2 pmol/L; p < 0.001). Total and intact GLP‐1 clearance was higher for subcutaneous than intravenous GLP‐1 ( p < 0.001 and p = 0.002, respectively). The increase in insulin secretion was greater, and glucagon was suppressed more with intravenous GLP‐1 ( p < 0.05 each). Gastrointestinal symptoms did not differ ( p > 0.05 each). Conclusions Subcutaneous GLP‐1 administration is much less efficient than intravenous GLP‐1 in lowering fasting plasma glucose, with less stimulation of insulin and suppression of glucagon, and much less bioavailability, even at fourfold higher infusion rates.