The explanation that travels with these drugs is that they slow the stomach, so a smaller meal feels like a larger one. That claim has been measured on the drug itself, and at twenty weeks it is gone. In 72 adults with obesity randomized to once-weekly semaglutide 2.4 mg or placebo, gastric emptying at week 20 showed no effect on the first-hour measure, and the five-hour measure lost significance once week-20 body weight was accounted for (P = 0.12).[1] At that same visit, the treated group ate 35% less at a free lunch — 1736 against 2676 kJ — and had lost 9.9% of body weight against 0.4% on placebo.[1] The stomach effect had faded. The eating effect had not.
A hormone found doing something else entirely
GLP-1 was not discovered as an appetite hormone. It was isolated inside the incretin effect: the observation that glucose taken by mouth provokes far more insulin than the same blood glucose delivered intravenously, because the gut releases hormones that the vein does not. Fourteen adults with type 2 diabetes and eight matched healthy controls were given a 50 g oral glucose load and, separately, an isoglycemic intravenous infusion. Gut hormones accounted for 72.8 ± 6.9% of the insulin response in the healthy group and 36.0 ± 8.8% in the diabetic group (P ≤ 0.05).[2] That halving is why the class exists, and it is a glucose finding. Weight came later, and the two indications still diverge, which is the subject of the two-approvals article.
The order matters for reading a sales page. A drug developed to restore an insulin signal turned out to reduce food intake, and the appetite effect was characterized after the molecule already existed, not before. That is the opposite of the usual story, in which a mechanism is worked out and a drug is then designed against it. It also means the appetite pathway was mapped in animals afterward, to explain a human result that had already been observed — which is why the mechanistic literature below is largely rodent and the efficacy literature is entirely human.
The hormone and the drug are not the same object
Native GLP-1 is destroyed within minutes of release, so the only way to study it in people is to infuse it continuously. Pooled subject-level data from 115 adults across intravenous infusion studies found energy intake at the following meal reduced by 727 kJ (95% CI, 548 to 908), or 11.7%. The absolute reduction was larger in lean participants (863 kJ) than in overweight ones (487 kJ), though the relative reduction was similar (13.2% against 9.3%), and infusion rate was the only independent predictor of the effect (r = 0.4, P < 0.001).[3]
One detail in that pooled analysis is worth more than the headline. The rise in plasma GLP-1 correlated with how full and how unwilling to eat more people said they felt — prospective consumption r = 0.40, fullness r = 0.38 — and did not correlate with how much they actually ate.[3] An eleven-percent cut in a single meal from an infusion of the natural hormone sits beside a 14.9% reduction in body weight over 68 weeks from a weekly injection of an engineered analog.[4] The drug is not the hormone turned up. It is a molecule redesigned to persist for a week, at exposures the gut never produces.
Gastric emptying, and how fast it stops mattering
The deceleration is real, and it is measurably temporary. Nine healthy volunteers received intravenous GLP-1 at 0.8 pmol/kg/min or placebo over 8.5 hours, with two liquid mixed meals four hours apart. Emptying was slowed significantly more after the first meal than after the second (P = 0.01). Pancreatic polypeptide, a marker of vagal activation, fell after the first meal and not the second, and glucose declined after meal one but rose after meal two.[5] Tachyphylaxis inside a single day, at the level of the nerve.
Longer exposures repeat the pattern. Thirty adults with obesity took semaglutide escalated to 1.0 mg for 12 weeks in a crossover trial: first-hour emptying was delayed, at an estimated treatment ratio of 0.73 (95% CI, 0.61 to 0.87), while overall emptying across five hours was not statistically different from placebo.[6] At 2.4 mg and twenty weeks, neither window showed an effect.[1] The honest summary is that delayed emptying is an early, first-meal phenomenon that attenuates, that it explains postprandial glucose better than it explains a year of weight loss, and that it explains most of the nausea — which is why nausea follows the same timeline and why the motility questions covered in the gastroparesis article are about a minority, not the mechanism.
Brain or gut: the evidence is animal, and it says both
Where the appetite effect is generated has been mapped in rodents, not in people. In mice and rats, semaglutide did not cross the blood-brain barrier. It reached the brainstem, the septal nucleus and the hypothalamus through the circumventricular organs and a handful of sites adjacent to the ventricles, and produced c-Fos activation in 10 brain areas — including regions such as the lateral parabrachial nucleus, where it has no direct receptor contact at all.[7] The drug does not bathe the brain. It touches a small number of doors and the signal propagates inward.
The gut arm has been isolated the same way. Using intersectional genetics in mice, gut-innervating vagal afferents that express the GLP-1 receptor were shown to relay meal-terminating signals to parabrachial neurons; activating them also improved glucose tolerance, and silencing them raised blood glucose independently of food intake. A separate afferent population marked by GPR65 drove hepatic glucose production and was dispensable for feeding altogether.[8] Two routes, one destination, both demonstrated in mice. No human experiment has separated them, and any consumer page that assigns a percentage to gut versus brain is assigning it to nothing.
The effect is on intake, not on burn
This is the part most often reversed in marketing copy. In the same 30-person crossover trial, semaglutide at 1.0 mg cut total ad libitum energy intake across lunch, dinner and snacks by 24% (−3036 kJ, P < 0.0001), lowered hunger and food cravings, and reduced the relative preference for high-fat foods — while resting metabolic rate adjusted for lean body mass did not differ between treatments. Weight fell 5.0 kg, predominantly as fat.[9] The rodent work reports the same direction: food intake and body weight fell without a decrease in energy expenditure.[7]
Nothing here speeds up a metabolism. The entire measured effect sits on the intake side of the equation, which is exactly why what comes off depends so heavily on what is eaten and whether anything is being trained — the concern in the body-composition article.
Four things the mechanism does not explain
Who responds. In STEP 1, 86.4% of participants on semaglutide reached a 5% reduction, 69.1% reached 10% and 50.5% reached 15%, against 31.5%, 12.0% and 4.9% on placebo.[4] Read the first figure the other way: roughly one in seven people on the full dose did not reach 5%, on the same receptor and the same titration. No feature of the pharmacology identifies them beforehand, and no test sold with a subscription does either. The placebo column is worth reading alongside it: 31.5% of people taking nothing reached 5% too, so a single person’s result is never clean evidence that the drug did or did not work for them.
Why it stops. Receptor agonism does not switch off at a particular week, yet the weight curve flattens — a separate problem covered in the plateau article.
Why it reaches things that are not food. The reward circuitry implicated in the rodent maps is not food-specific, and the clinical signals in alcohol and nicotine are much weaker than the mechanistic story predicts, which is the whole argument in the alcohol-use article.
Why the dose ladder works. Escalation exists because tolerability demands it, not because the receptor requires a run-up. The schedules are in the titration article, and the expected-loss arithmetic is in the expected weight loss tool.
What the mechanism does and does not certify at checkout
Every measurement above was made on FDA-approved product at labeled doses under trial conditions. Most sellers covered here dispense compounded semaglutide or tirzepatide, which is not FDA-approved and is not reviewed by the FDA for safety, efficacy or quality before it is dispensed. A shared mechanism is a claim about a molecule, not about a vial: it says nothing about what concentration arrived, and nothing about whether the dose on the label is the dose in the syringe.
The practical reading of all of this is narrow and useful. The drug works by making less food feel like enough, the stomach contributes early and then recedes, nothing about it raises the rate at which calories are burned, and about one person in seven at full dose will not get a meaningful result from it. A page that promises a metabolic reset is describing a mechanism that has not been measured, on a product whose contents have not been verified.