The reports arrive in two directions at once. Some people say they sleep harder than they have in years. Others say they surface at three in the morning and cannot settle, or that their dreams have become loud and strange. Both accounts are common enough to have their own threads, and the published evidence behind them is far smaller than the evidence behind the weight figures. Breathing during sleep is the separate, better-studied question, and it lives in the sleep-apnea article. This is about the rest of the night.
The one prospective sleep-quality dataset
A 2026 prospective observational study enrolled patients with obesity starting semaglutide or tirzepatide and measured sleep with the Pittsburgh Sleep Quality Index at baseline and at three months. It covered 78 patients — 52 on tirzepatide, 26 on semaglutide — with no control group.[1]
Median score fell from 6.0 to 3.0, and the share scoring above 5, the cutoff for poor sleep, fell from 55.1% to 16.7%, both at p < 0.001. Depressive symptom burden moved on the same clock: median questionnaire score from 10.0 to 6.0, with clinically significant burden falling from 55.1% to 11.5%.[1]
Then the finding that complicates the obvious reading. In multivariable models, percentage weight loss was not significantly associated with sleep response, and neither was which drug the patient took. What predicted a sleep response was a worse baseline score and a larger improvement in depressive symptoms.[1] The folk explanation — you sleep better because you weigh less — is the one thing this dataset did not find. Seventy-eight people over three months without a placebo arm cannot settle why, and the mood link is covered separately in the mental-health article.
The randomized measurement is smaller than it sounds
One trial put wrist-worn accelerometers on people taking this drug class. 195 adults with obesity completed an eight-week low-calorie diet, losing a mean of 13.1 kg, and were then randomized in a two-by-two design to a year of weight-loss maintenance with or without exercise and with liraglutide 3.0 mg daily or placebo.[2]
During maintenance, the liraglutide groups showed 5 minutes more sleep a night at 26 weeks against 15 minutes less on placebo — a difference of about twenty minutes — and the difference was gone at one year. The arm that held on to the sleep-quality improvement earned during the initial weight loss was the exercise arm, not the drug arm.[2] That is the entire randomized measurement of sleep on a GLP-1: one molecule, one accelerometer endpoint, one transient difference. The exercise finding is consistent with what the exercise article describes elsewhere.
Sleep predicts the weight result, not the other way round
The same trial ran the question backwards, and that is where the numbers get large. Participants who were short sleepers at randomization — under six hours a night — regained 5.3 kg during the one-year maintenance phase (p = .0008) and had less reduction in body fat percentage (p = .007) than participants sleeping normally. Participants whose sleep quality was poor before the weight loss began regained 3.5 kg more than good sleepers (p = .010).[2]
Read alongside the twenty-minute drug effect, the asymmetry is stark. Sleep moved the weight outcome by kilograms. The drug moved sleep by minutes, temporarily. For anyone budgeting for a year of treatment on the price boards, the cheaper variable is the one measured in hours a night.
Short sleep changes what comes off, not just how much
A randomized crossover study put 10 overweight adults through two 14-day periods of identical moderate caloric restriction, once with 8.5 hours of sleep opportunity a night and once with 5.5. Weight loss was similar. The composition was not.
Sleep curtailment cut the proportion of weight lost as fat by 55% and increased the loss of fat-free body mass by 60%, from 1.5 kg to 2.4 kg (P = 0.002), alongside increased hunger and a shift in substrate use away from fat oxidation.[3] Ten people for two weeks is a very small experiment, and it is the cleanest one available: the same deficit, deliberately manipulated sleep, opposite tissue outcomes. It is also the clearest reason the lean-mass question in the muscle article is not purely a drug question.
The hormonal mechanism was described earlier still. Sleep curtailment in healthy young men was associated with decreased leptin, elevated ghrelin, and increased hunger and appetite.[4] Appetite suppression is precisely what these drugs are bought for, and short sleep pushes the same dial the other way.
Insomnia: what the safety record actually contains
Insomnia has been measured as an outcome in one large matched cohort. A 2026 retrospective study of electronic health records compared adolescents with overweight or obesity starting semaglutide, liraglutide or tirzepatide against those receiving lifestyle intervention, propensity matched to 9,222 in each group, with outcomes assessed from 30 to 1,095 days.[5]
No significant difference was observed for insomnia, nor for depression, suicide attempt or eating disorders. Suicide-related events were less frequent on drug (hazard ratio, 0.74; 95% CI, 0.57 to 0.95), as were suicidal ideation (0.73; 0.56 to 0.96) and anxiety (0.92; 0.84 to 0.99).[5] The subgroup pattern is worth keeping: the reductions were significant among female adolescents (0.61; 0.46 to 0.82) and among those without type 2 diabetes (0.63; 0.47 to 0.85), and not among males or among those with diabetes. Its authors also note that the difference did not hold against an active metformin comparator, which is the caution that keeps this from being a benefit claim.
Spontaneous reporting says something different and proves less. A FAERS analysis of 211,195 cases from 2021 to 2025, compared against other antidiabetic and anti-obesity agents, found 16 preferred terms meeting signal criteria, with suicidal ideation the most frequently reported at a reporting odds ratio of 2.95, and signal magnitudes consistently higher for semaglutide than tirzepatide.[6] A reporting odds ratio is a measure of what gets reported, not of what happens; the same analysis found that being 65 or older lowered the predicted probability of a psychiatric report, which is a statement about who files reports.
Where the sleep instruments did move in a trial
The one randomized program that administered validated sleep instruments enrolled people with moderate-to-severe obstructive sleep apnea and obesity. At week 52, tirzepatide significantly improved PROMIS Sleep-Related Impairment and PROMIS Sleep Disturbance scores against placebo, alongside functioning and quality-of-life measures.[7]
Two limits keep that from answering the general question. The participants had a diagnosed breathing disorder, so the mechanism available to them is not the mechanism available to a person who sleeps badly for other reasons. And in the arm already using positive airway pressure, that therapy was withdrawn before the assessments, which changes what the comparison is against.[7]
The vivid dreams have no literature at all
Dream content has never been an endpoint in this drug program, and no pharmacovigilance study has published a disproportionality figure for it. That is an absence of evidence rather than evidence of absence, and it cuts both ways: nobody has shown these drugs change dreams, and nobody has looked hard enough to say they do not. The plausible confounders are ordinary — disrupted sleep surfaces more dream recall, and so does waking more often — and the trial record contains no measurement that would separate them. What the trials did record on those nights is gastrointestinal, which is the subject of the reflux article and of the fatigue article.
What a buyer can take from this
Three things hold. Sleep quality does appear to improve on treatment, in one uncontrolled cohort of 78 people, and it did not track how much weight came off. Insomnia has been measured once at scale and was no more common than under lifestyle care. And the largest sleep effect in this entire literature runs the other way — short or poor sleep before treatment predicted kilograms of regain and a worse split between fat and lean.
Every drug figure above came from branded product at labeled doses in a trial or a coded medical record. Compounded drugs are not FDA-approved, and are not reviewed by the FDA for safety, efficacy or quality before they are dispensed, which is set out in the compounding article; no sleep measurement of any kind has been published on a compounded preparation. How a figure here is established before publication is described in the methodology.