Smell is the sense that decides what food is worth eating, so a drug that changes how much people want to eat invites the suspicion that it has changed how things smell. The evidence pointed at that suspicion is real but small, and it splits cleanly by instrument. Where the question was put to a reporting database, the answer came back that something is wrong. Where the question was put to an olfactory test, the answer came back that smell had improved. Both results exist, both are published, and nothing so far has reconciled them. The related complaint about flavor rather than odor belongs to the taste article, which reaches a similar destination by a different route.
Six labels count one chemical sense and not the other
The current prescribing information for WEGOVY, OZEMPIC, RYBELSUS, ZEPBOUND, MOUNJARO and SAXENDA contains no olfactory term at all. Not smell, not olfactory, not anosmia, parosmia, hyposmia, dysosmia or phantosmia — zero occurrences across all six documents.[1][2][3]
A zero is only worth reading if something comparable in the same corpus is not zero, and here something is. Dysgeusia appears in every one of the six. The WEGOVY label records it in 1.7% of adults treated for weight reduction against 0.5% on placebo.[1] ZEPBOUND records it in 0.4% of treated patients and in no placebo patient.[2] The OZEMPIC label groups it with fatigue and dizziness among reactions occurring at a frequency above 0.4%. RYBELSUS lists it under nervous system disorders in its postmarketing section — the part of a label reserved for effects reported spontaneously after approval, where no trial denominator exists and the bar for inclusion is lowest.
That last detail is the one that carries weight. A smell complaint would have had a place to land in these documents even without a trial behind it, in exactly the section where the taste complaint landed, and it did not land there. These labels are not silent about chemosensory effects. They are silent about this one.
The reporting databases are not silent
A pharmacovigilance analysis of the FDA Adverse Event Reporting System assembled 25,110 neuropsychiatric adverse-event reports involving GLP-1 receptor agonists and ran disproportionality testing across eight categories. Olfactory nerve abnormalities returned a reporting odds ratio of 2.44 (95% CI, 1.83 to 3.25) and sensory nerve abnormalities 1.69 (1.54 to 1.85), against headache at 1.74 (1.65 to 1.84) and migraine at 1.28 (1.06 to 1.55). The authors described the olfactory signal as detected for the first time in that analysis. Median time to onset across the whole set was 16 days, interquartile range 3 to 66.[4] The headache and migraine figures from the same table are handled in the migraine article.
A separate query of the same database, restricted to otolaryngologic events across exenatide, liraglutide, dulaglutide, semaglutide and tirzepatide, collected 9,746 reports and required a proportional reporting ratio of 2 or above with a lower confidence bound on the reporting odds ratio exceeding 1. Anosmia reached significance for semaglutide and for no other agent in the set. Liraglutide and exenatide cleared the same threshold for dysgeusia but not for anosmia.[5]
Neither result is an incidence. A disproportionality ratio says a term occupies more of the database than its background share predicts; it cannot say how many people out of a hundred experienced it, and it rises with prescribing volume, publicity and the length of time a drug has been on the market. The distinction is the same one that governs every signal in the tolerability picture, and it matters more here than usual, because the signal is the entire case for the effect existing at all.
Every prospective measurement found smell improving
This is where the direction reverses. A clinical and neuroimaging study in China assessed cognition, olfactory function and odor-induced brain activation in 35 people with obesity and type 2 diabetes, 35 people with type 2 diabetes without obesity, and 35 matched controls. Twenty of the participants with obesity and inadequate glycemic control then received a GLP-1 receptor agonist for three months and were reassessed. Their olfactory test total score improved, alongside a higher Montreal Cognitive Assessment score and enhanced odor-induced activation of the right parahippocampus.[6]
Twenty people, no control arm on the treatment phase, and a population with type 2 diabetes rather than the cash-pay weight-loss population — all three limits apply. What makes the result hard to dismiss is that a scoping review reached the same conclusion from a different corpus. Six databases were searched for treatments of olfactory dysfunction in type 2 diabetes; 3,631 articles were identified and three met inclusion. Of the interventions ever tested in that setting — hyperbaric oxygen, the DPP-4 inhibitor linagliptin, and the GLP-1 agonists exenatide and liraglutide — only hyperbaric oxygen and the GLP-1 agonists produced statistically significant improvements in olfactory identification.[7]
Three articles out of 3,631 is a thin literature by any standard, and the review said so. But it is a literature that points one way, and the way it points is up.
The weight loss itself confounds the question in the same direction
Obesity is associated with worse olfaction at baseline, which means anyone losing a large amount of weight has room to improve regardless of the drug that caused it. A systematic review screened 11,311 studies and included 30 covering 2,482 participants. Fourteen of the 21 non-interventional studies found impaired smell capacity in people with high body weight. Nine studies examined bariatric surgery, and six of the nine reported significant postoperative improvement in olfaction. Two studies measuring olfactory bulb volume found a significant correlation with body-mass index, though the review called the bulb-volume findings discordant.[8]
Bariatric surgery is the sharpest available comparator for rapid weight loss without a GLP-1 receptor agonist, and it improved smell in two-thirds of the studies that looked. The surgical and pharmacological routes are compared on their other outcomes in the surgery comparison. For this question the implication is narrow and uncomfortable: a person on one of these drugs is simultaneously exposed to a suspected cause of olfactory harm and to the single best-documented cause of olfactory improvement, and no published design separates them.
The one case where both were measured
A 41-year-old woman received semaglutide for weight management for 24 weeks with olfactory assessment performed before and after, using a validated Questionnaire of Olfactory Disorders, the Yonsei Olfactory Function Test as a psychophysical measure, and chemical gustometry. At baseline she had no olfactory complaints and normal results on all three.
After 24 weeks she reported parosmia — distorted rather than absent smell — and scored her olfactory discomfort at 48 on a visual analog scale with a maximum of 50, with measurably reduced quality of life. Her Yonsei Olfactory Function scores were stable, and her gustatory function was normal. The authors titled the report around the discrepancy itself.[9]
One patient establishes nothing about frequency. What it does establish is that the two instruments can disagree inside a single person, prospectively measured, which is the cleanest demonstration available that the symptom-versus-test split is not an artifact of comparing different populations. A person can be genuinely and severely bothered by what they smell while a psychophysical test records no deficit. That pattern is familiar from the subjective-objective gap documented for dryness in the dry mouth article, and it means a normal smell test is not a reason to tell someone nothing is happening.
Where a mechanism would have to live
The preclinical work that touches olfactory tissue does not predict a deficit either. In lean mice, regional blood-brain barrier transport rates for a radiolabeled dulaglutide fragment varied roughly 2.5-fold across brain regions, and the fastest transport of all was into the olfactory bulb, followed by frontal cortex, cerebellum and pons.[10] A systematic review of neurogenesis, drawing on animal models across exenatide, geniposide, liraglutide, lixisenatide and semaglutide, found that GLP-1 receptor agonists increased neurogenesis in the dentate gyrus, hippocampus, olfactory bulb and medial striatum.[11]
Those are rodent findings and they carry every limit rodent findings carry. Read at face value they say the olfactory bulb receives more of the drug than most of the brain and responds by growing new neurons, which is a mechanism for smell getting better rather than worse. A 2026 review of GLP-1 receptor agonists in otolaryngology notes that the receptor participates in sinopulmonary inflammatory cascades and that recent evidence suggests clinical implications for both chronic sinusitis and olfactory disorders, without resolving the direction.[12]Nasal inflammation is the obvious candidate for a peripheral route, and it has not been measured in anyone taking these drugs.
What a compounded vial changes about this
None of the measurements above involved a compounded product. Compounded semaglutide and tirzepatide, which is what most of the sellers listed in the semaglutide board dispense, have not been approved by the FDA, and the agency does not review them for safety, effectiveness or quality before a pharmacy sends one out. The reporting analysis that covers them examined 81,078 GLP-1 reports, of which 707 involved compounded products, and found elevated reporting odds for abdominal pain at 2.84 (95% CI, 2.29 to 3.49), diarrhea at 1.59 (1.25 to 1.99) and nausea at 1.27 (1.05 to 1.52). No olfactory term is among the outcomes it reports. The same analysis found reporting odds of 48.92 (12.63 to 189.6) for preparation errors and 19.00 (4.24 to 85.03) for contamination.[13] A term a study never examined is not a term a study found to be absent, and the quality figures are the part of that record with a clear bearing on a purchase.
What can honestly be said
The defensible summary is short. No label counts this. The reporting databases produce a genuine disproportionality signal, strongest for semaglutide and strongest for distorted rather than absent smell, with a median onset around two weeks. Every prospective measurement of olfactory function under one of these drugs found it improved, in populations that had room to improve, and the weight loss alone would predict the same direction. The single case with before-and-after objective testing found a severe symptom and an unchanged test.
What that adds up to for a person deciding whether to start is that a distorted sense of smell is an uncommon and plausible complaint, that it is likely to arrive early if it arrives, and that nothing published suggests a lasting loss of olfactory capacity. A complete or sudden loss of smell is a different presentation with its own causes — nasal obstruction, infection, head injury, neurological disease — and it earns a workup rather than an assumption that the injection did it. Smell change arriving together with facial numbness, weakness or a new headache pattern belongs with a prescriber promptly, and the seller-level questions that sit underneath any of this are in the provider write-ups.