Cozy Butter · Free tool · How the medication works
Your GLP-1 medication is doing something interesting in your brain, and it is probably not working the way you have heard it explained. The drug does not cross into the brain. It reaches a short list of places where the barrier has an opening, and much of what happens afterwards happens second hand. Follow the signal below, from your own natural GLP-1 through to a dose of medication, and see which parts are actually involved.
Tap any area to read what it does and what was measured there. The columns group areas by when the drug was found there rather than by where they sit in the brain, and the list above is grouped by anatomy. The solid arrow into the parabrachial nucleus is the relay the authors measured from the brainstem. The dashed one is the additional route they propose from the hypothalamus, which is a hypothesis rather than a measurement.
The GLP-1 your gut makes has a half-life of under two minutes, because DPP-4 starts breaking it down almost immediately. That is too short a window to travel through the bloodstream and act on the brain in any quantity. What it does instead is act locally, on GLP-1 receptors sitting on the endings of the vagus nerve in the gut and liver area, and the nerve carries the message up to the brainstem.
Your brain also produces GLP-1 of its own, in a group of neurons inside the nucleus tractus solitarius. Those neurons project widely, and they are the likely route by which GLP-1 signalling reaches receptors in areas a circulating drug cannot get to.
Semaglutide and the other agonists are built so DPP-4 cannot cut them, which is why they last long enough to be taken once a week. That does not buy them entry to the brain. When researchers tagged semaglutide and imaged where it went, it was not in the brain generally. It was at the four circumventricular organs, which are the small areas where the blood-brain barrier is open by design so the brain can sample the blood, and at a handful of sites lining the ventricles nearby.
A review of how incretin drugs get to their targets makes the same point: the receptors that matter mostly sit behind the barrier, the newer long-acting agonists do not cross it, and they still manage to reach a few useful sites. How they do that is an open research question rather than a settled one.
The nausea is not a stomach side effect that happens to come along with an appetite effect. Both start in the same small region of the brainstem. What researchers did not expect is that they start in different populations of neurons within it, which can be switched on separately.
Nucleus tractus solitarius
In mice, the GLP-1 receptor neurons in the nucleus tractus solitarius responded mainly to nutrients. Switching them on produced fullness and reduced eating, and it did not produce aversion. This is the effect people are hoping for when they start a medication.
Area postrema
The GLP-1 receptor neurons a few millimetres away in the area postrema responded to almost anything. Switching them on produced strong aversion alongside the reduction in eating. The area postrema is the brain's poison detector, and a circulating drug reaches it because there is no barrier in the way.
Current medications reach both, which is why nausea and appetite suppression arrive together. In the trials behind the Wegovy label, 44 percent of people reported nausea compared with 16 percent on placebo. The finding that the two circuits can be separated is the reason researchers now think a medication could suppress appetite without the nausea, and that work is at the animal stage.
There is a related result behind tirzepatide, which acts on the GIP receptor as well as the GLP-1 receptor. In mice, rats and musk shrews, switching on the GIP receptor blocked vomiting and reduced illness behaviour caused by GLP-1 receptor activation, while the reduction in food intake and body weight held. The GIP receptor turns out to be expressed in the same brainstem region, largely on inhibitory neurons. Whether that is why some people tolerate tirzepatide differently has not been tested head to head in people.
Forty-eight people, some lean and some with obesity or type 2 diabetes, were scanned while looking at pictures of food. On exenatide rather than placebo they ate less, and the response to those pictures dropped in the insula, the amygdala, the putamen and the orbitofrontal cortex. Those are appetite and reward areas.
In the same study, giving a GLP-1 receptor blocker first largely cancelled the change. That is what makes this stronger than an observation: the effect went away when the receptor was blocked, so it was the receptor producing it rather than the weight loss or the eating.
A separate study blocked GLP-1 receptors while people ate a real meal. Normally, eating reduces the insula's response to food pictures. With the receptor blocked, that reduction did not happen in the people with type 2 diabetes. The drug is amplifying a mechanism you already have rather than adding a new one.
Nobody has imaged where semaglutide physically goes inside a living human brain. The access map and the firing map are both from mice, and the human evidence is functional: what changes in a scanner, what people eat, and what they report. Both kinds are real, and they answer different questions.
The ventral tegmental area and the nucleus accumbens both carry GLP-1 receptors, and the labelled drug was not found in either of them under any dosing condition. If a medication changes how rewarding food is, the change is arriving through connections from the areas it does reach. It is not the drug acting on the dopamine system directly.
Food restriction activates the brain's own GLP-1-producing neurons in the brainstem. A GLP-1 agonist given by injection does not, and chronic dosing lowered the production of that peptide instead. The medication works alongside your own system rather than by restarting it.
Observational data had suggested that people on GLP-1 medications developed dementia less often. Two phase 3 trials tested it properly, in 3,808 people with early Alzheimer's disease. Oral semaglutide did not slow decline on the primary measure at two years in either trial, and the planned one-year extension was cancelled.
A lot of people on these medications report that wanting things other than food got quieter too, and alcohol comes up most often. There are now three randomised trials, and the largest is the one that settles the most. Over 26 weeks, 108 adults with alcohol use disorder and obesity received the full 2.4 milligram weekly dose or placebo, with cognitive behavioural therapy offered to both groups. Heavy drinking days fell by 13.7 percentage points more on semaglutide than on placebo, which was the measure the trial was built to answer. Total alcohol drunk, drinks per drinking day, craving, and blood markers of drinking all came down with it.
The two smaller trials are more mixed. In one, 48 adults took a low dose for nine weeks and drank less in a laboratory session and on the days they did drink, while average drinks per day and number of drinking days did not move. In the other, 50 adults took oral semaglutide for eight weeks and the main measure, craving triggered by alcohol cues in a laboratory, came out no different from placebo, though drinks per drinking day, craving outside the laboratory and alcohol-related consequences all improved.
So there is an effect, and it is clearest at the full dose over the longest run. What none of these trials tested is whether it holds after the medication stops, or whether it works the same way in people who do not also have obesity.
The animal work points at the reward areas the drug cannot reach. When a GLP-1 agonist was injected straight into the ventral tegmental area, the nucleus accumbens or the lateral hypothalamus of rats, alcohol intake fell and so did the effort the rats would put in to obtain sugar. Injecting it into the arcuate nucleus did not change alcohol intake, though it did reduce the work for sugar.
That sits awkwardly next to the map, and the awkwardness is the honest part. In animals the sites that matter most for craving are ones an injected medication never reaches, and in people the effect shows up anyway. However it resolves, it is not the drug acting on the dopamine system directly.
In the STEP 1 extension, 228 people who had finished 68 weeks of semaglutide stopped the medication and were followed for another year. They had lost 17.3 percent of their body weight on average, and they regained about two thirds of it. Most of the cardiometabolic improvements moved back toward where they started as well.
That result makes sense given everything on this page. The medication supplies a signal from outside to a brainstem that keeps asking for one. It does not repair the signals the body reduced, and it does not retrain the circuit. When it is withdrawn, the circuit goes back to running on the body's own inputs, which is why the trial authors describe obesity as a chronic condition requiring ongoing treatment. The food noise simulator covers what those underlying signals do when weight comes off.
The whole-brain access map and the activation map come from one study, which imaged fluorescently labelled semaglutide and counted activated neurons, both in mice. Rodent and human brains share this machinery closely enough for the work to be the basis of current drug development, and they are not identical. Nothing here has been measured inside a living human brain, and this page does not claim otherwise.
The map is also specific to semaglutide. Liraglutide reaches a slightly different set of areas in the same experiments, and tirzepatide acts on a second receptor that has not been mapped the same way. Nothing on this page calculates anything about you, and the areas the drug reaches are not areas you can feel.
This page explains published research. It does not diagnose anything and it does not measure anything about you. It is also not a reason to change a medication on your own, because that decision belongs with your prescriber.