Every telehealth intake mentions exercise, and almost none of them say what kind, how often, or on what evidence. The useful question is narrow: has anyone randomized people to training alongside one of these drugs and measured what it added? One trial has, and its answer is more specific than the advice usually attached to a semaglutide subscription.
The one trial that randomized both
A Copenhagen group enrolled adults with a body-mass index of 32 to 43 and no diabetes. All of them first went through an eight-week low-calorie diet, which took off a mean of 13.1 kg. Then 195 were randomized to one of four year-long strategies: a moderate-to-vigorous exercise program with placebo, liraglutide 3.0 mg daily with usual activity, both together, or neither.[1]
Against placebo at one year, the exercise arm held 4.1 kg more (95% CI, −7.8 to −0.4), the liraglutide arm 6.8 kg more (95% CI, −10.4 to −3.1) and the combination 9.5 kg more (95% CI, −13.1 to −5.9).[1] Training alone beat doing nothing. The drug alone beat training. Adding training to the drug beat the drug, though the 2.7 kg gap between combination and liraglutide alone did not reach significance on its own.[1] That last interval is the one to keep in view. It is the comparison a coaching upsell is implicitly quoting, and it crosses zero.
The weight number is not the interesting one
Body-fat percentage was a prespecified secondary endpoint, and it separated the arms more cleanly than weight did. The combination cut body-fat percentage by 3.9 percentage points, about twice the reduction in the exercise arm (−1.7 points; 95% CI, −3.2 to −0.2) and about twice that in the liraglutide arm (−1.9 points; 95% CI, −3.3 to −0.5).[1]
Only the combination improved glycated hemoglobin, insulin sensitivity and cardiorespiratory fitness.[1] Two arms lost similar amounts of weight and came out in measurably different physical condition, which is the whole argument for reading a scale as an incomplete report. What the scale conceals is the subject of the body-composition article.
A year after everyone stopped
The same cohort was reassessed 12 months after all treatment ended. Of those invited, 109 attended. Compared with people coming off liraglutide alone, people coming off the combination were 5.1 kg lighter (95% CI, −10.0 to −0.2) with a body-fat percentage 2.3 points lower (95% CI, −4.3 to −0.3).[2]
Odds of still holding a 10% loss a year after stopping ran 7.2 times higher for the former combination group than for former placebo (95% CI, 2.4 to 21.3). Against former liraglutide alone the figure was 4.2 (95% CI, 1.6 to 10.8).[2] Regain over that year was 6.0 kg larger after stopping liraglutide than after stopping supervised exercise.[2] A habit and a prescription do not decay at the same rate, which matters when the prescription ends, as the withdrawal data describes.
Which kind of training, from a trial with no drug in it
The Copenhagen protocol was a mixed program, so it cannot say whether lifting or cardio did the work. The closest answer comes from a trial with no medication at all: 160 adults over 65 with obesity, all dieting, randomized to aerobic training, resistance training, both, or a control group.[3]
Weight fell by about 9% in every exercise group. Lean mass fell 5% in the aerobic group, against 3% in the combined group and 2% in the resistance group. Strength rose 19% with resistance training and 18% with the combination, against 4% with aerobic training alone, while peak oxygen consumption rose most in the arms that included aerobic work.[3] Total-hip bone density fell 3% in the aerobic group against 0.5% with resistance training.[3]
That is an older, non-diabetic, drug-free population dieting for six months. It is evidence about what different training does during weight loss, not evidence about GLP-1 receptor agonists, and anyone presenting it as the latter is stretching it.
Bone is the other tissue in this argument
A prespecified secondary analysis of the Copenhagen trial scanned hip, spine and forearm. Bone density in the combination group was statistically indistinguishable from placebo at the hip and spine despite a mean weight loss of 16.88 kg. Liraglutide alone lost more hip density than exercise alone (−0.013 g/cm², 95% CI, −0.024 to −0.001; P = .03) at a similar weight loss.[4]
The comparison worth noticing is that last one. Two groups came down by comparable amounts and their skeletons did not respond identically. Whether that finding transfers to semaglutide or to the molecule priced on the tirzepatide board has not been tested.
The trial that would settle it has not reported
A protocol published in 2026 describes a 232-participant randomized trial at a single center in Kuwait, assigning adults starting semaglutide or tirzepatide to control, home-based resistance exercise, protein supplementation, or both. The exercise arm prescribes three sessions a week progressing from one to three sets, and the primary outcome is quadriceps cross-sectional area on MRI.[5]
Until that reports, the honest summary is this. Exercise plus a GLP-1 has been randomized once, with liraglutide, and it improved body composition, fitness and post-treatment durability beyond the drug alone. Which training, at what dose, alongside the two molecules people actually buy on the seller index, is inference rather than result.
What none of this authorizes
Nothing above is a program for a particular person. Exercise-related adverse events in the older-adult trial included musculoskeletal injuries, and the participants in both trials were screened, supervised and progressed by staff.[3] What training suits someone carrying a joint problem, a cardiac history or a rapid rate of loss is a conversation with a prescriber or a physiotherapist, not a paragraph.
One commercial note follows from the evidence rather than the marketing. A plan that bundles coaching is charging for something the trials treated as the comparator, not the intervention, and a plan that bundles nothing is not thereby worse. The way a claim on this site gets established before it is published is set out in the methodology.