Search for coffee and these drugs and the results arrive with confidence: skip the morning cup, caffeine makes the nausea worse, it dehydrates you when you are already eating less, space it hours from the injection. None of that comes from a study. No trial has given caffeine to anyone taking semaglutide, tirzepatide, liraglutide, dulaglutide or exenatide and measured what happened — not a small trial, not an inconclusive one. The combination is unstudied, which is a different and more useful thing to know than a guess. What is established sits either side of the gap, and it is worth setting out before reaching for the advice, much as the drug interaction record turns out to be narrower than its reputation.
Caffeine is unusually sensitive to the stomach
This much has been measured, and early. A 1988 study compared caffeine absorption in patients with gastric stasis, in patients after partial gastrectomy with adequate emptying, and in healthy subjects given liquid test meals to slow them down. In gastric stasis the absorption rate constant fell to 0.018 per minute against 0.122 in controls, and time to peak stretched to 160 minutes against 46. Peak concentration fell to 5.9 micrograms per milliliter against 17.7.[1]
Both liquid test meals reproduced the pattern in healthy people, pushing time to peak to roughly 91 and 93 minutes. Patients whose emptying was adequate after surgery absorbed at a normal rate. Slow the stomach and caffeine arrives later and lower; that relationship is not in question.
The relationship is so tight that caffeine became the instrument
Pharmaceutical scientists now use salivary caffeine kinetics as a gastric emptying measurement. A 2018 validation gave six volunteers 35 mg of caffeine in an ice capsule with 240 mL of water, fasted and after a high-fat breakfast, and compared saliva sampling against magnetic resonance imaging in the scanner. Normalized, the two methods tracked each other with a mean correlation coefficient of 0.912 fasted and 0.887 fed.[2]
That is the closest thing to a finding this page can offer. Because caffeine is a validated readout of gastric emptying, and because these drugs demonstrably delay gastric emptying — the same property that puts a stomach at issue before anesthesia — the direction of the effect can be stated with confidence. Caffeine should peak later and lower during treatment, most sharply after a first dose or a dose increase. The size of that shift is unknown, because the tracer has never been run on anyone taking one.
The one human measurement points the other way
One clinical study has measured caffeine exposure alongside an incretin-based peptide. NN1177, a glucagon and GLP-1 receptor co-agonist, was assessed in 45 participants who received the Cooperstown 5+1 cocktail — midazolam, caffeine, omeprazole, dextromethorphan and S-warfarin — alone and again at steady-state drug exposure. Midazolam and S-warfarin exposure were unaffected. Omeprazole, caffeine and dextromethorphan generally showed decreases in both total exposure and peak concentration.[3]
Two cautions attach. NN1177 is not semaglutide or tirzepatide, and a co-agonist that also hits the glucagon receptor is a different drug with different metabolic effects. And the study existed to test a laboratory-derived worry about enzyme suppression, which it disproved — the in vitro downregulation that prompted it did not appear in people at all. But it is the only human caffeine measurement in this drug family that exists, and it shows less caffeine exposure rather than more. The popular advice assumes the opposite.
There is a general lesson in how that study came about. Laboratory work on isolated human liver cells had shown the peptide suppressing cytochrome P450 enzymes, and a simulation built on that finding flagged caffeine and midazolam as safety concerns. The clinical study was run to check, and found no such effect — the authors conclude that standard in vitro systems predict interactions poorly for peptide drugs and that better preclinical models are needed.[3] A mechanism that survives a model and dies in a person is the normal outcome here, which is worth remembering before treating any confident mechanistic claim about caffeine and these drugs as settled.
Coffee does not do what the advice assumes to appetite either
A second assumption sits underneath: that coffee blunts appetite, so pairing it with an appetite-suppressing drug compounds something. That has been tested directly. A double-blind randomized crossover gave 12 adults four conditions — placebo, decaffeinated coffee, caffeine, and caffeine with decaffeinated coffee — after a breakfast labeled with carbon-13 octanoic acid, with breath samples for gastric emptying and an ad libitum meal four and a half hours later.[4]
Energy intake at that meal was 2118, 2128, 2287 and 2016 kilojoules across the four arms, with no significant difference (p > 0.05). Appetite sensations, plasma glucose and gastric emptying showed no significant differences either.[4] A 2017 review of the wider literature concluded that coffee taken 3 to 4.5 hours before a meal has minimal influence on intake, that caffeine taken 0.5 to 4 hours before may suppress acute intake, and that the evidence on gastric emptying, appetite hormones and appetite perception is equivocal.[5]
Weight loss changes drug metabolism, but not caffeine's enzyme
There is one more route by which a GLP-1 could plausibly change how caffeine behaves: caffeine is cleared by CYP1A2, and losing a large fraction of body weight changes enzyme activity. That has been measured, in weight loss achieved by other means.
A 2025 study gave a cocktail of acetaminophen, caffeine, omeprazole and midazolam to patients before and after a structured weight-management program. Among the 12 who completed both phases, after weight loss of 7.6% to 26.2%, CYP2C19 activity rose by a median 90.5% and CYP3A4 by 43.0% (P < .05). CYP1A2 was measured in the same cocktail and was not among the enzymes that moved significantly.[6]
A larger trial reached the same conclusion from a different direction. In 40 patients undergoing gastric bypass and 41 losing weight by diet, followed for two years with four pharmacokinetic investigations, CYP2C19 activity rose and stayed elevated after surgery, while no clinically significant effect on CYP1A2 or CYP2C9 was observed. Baseline CYP2C19 activity was 2.7-fold higher in normal-to-overweight controls than in patients with obesity.[7]
So the enzyme that clears caffeine appears to be the one weight loss leaves alone. Whatever changes about a cup of coffee during treatment is more likely to be about when it arrives than about how long it lasts.
What a question this common looks like when it is unstudied
The shape of this gap is worth naming, because it recurs. A question can be asked constantly and still have no trial behind it, and the volume of the asking exerts no pressure on the research agenda. Interaction studies get run on drugs a regulator requires or a sponsor worries about: the manufacturers tested metformin, warfarin, digoxin, statins, lisinopril and contraceptives, because those are what a trial population was already taking and what a label has to speak to. Nobody has to fund a caffeine study, and so nobody has.
That is different from a question having been answered in the negative. The four decades of gastric emptying work above make the direction of an effect predictable, and the one adjacent human measurement makes its magnitude genuinely uncertain. Anyone writing a confident rule is filling that space with mechanism, which is exactly the move that produced a prediction of enzyme suppression the clinic could not reproduce.
What a reader can reasonably do with this
The honest summary is short. The direction of the absorption effect is predictable and its size is unmeasured. The single human caffeine measurement in this drug family found exposure falling rather than rising, in a molecule nobody is prescribing. Coffee's own effect on appetite and emptying is small and inconsistent in healthy people, and nobody has looked at it in people on these drugs at all.
That leaves the practical questions to ordinary sources. Caffeine on an empty stomach is uncomfortable for some people regardless of what else they take, and nausea in the first weeks of treatment has its own well-characterized course in the tolerability evidence. Fluid intake matters more than usual when appetite is suppressed, which is also why a vomiting illness is handled differently on these drugs. Neither of those facts requires a rule about coffee, and none of the confident rules circulating online has a trial behind it.
One boundary is worth stating plainly, since this market sells the drug without a label attached. Compounded semaglutide and tirzepatide are not FDA-approved, and are not reviewed by the FDA for safety, efficacy or quality before they are dispensed — so there is no prescribing information to consult on any of this, and nothing that could be silent about caffeine in an informative way. Where a claim on this site comes from is described in the methodology.