Caffeine's half-life in adults is 3–7 hours, so half of an afternoon cup is still in you hours later. At a 5-hour half-life, an 8 fl oz mug of black tea (about 47 mg) drunk at 3 pm leaves about 16 mg at 11 pm. In the best-known timing study, 400 mg of caffeine taken 6 hours before bed cut measured sleep by more than an hour in 12 adults; a 2025 trial in 23 young men found no measurable effect from 100 mg taken 4 hours before bed.
This article reports what the studies measured: who took part, how much caffeine, and when. It is not medical advice. For caffeine per cup by tea type, see caffeine in tea vs coffee; for how brew time changes it, see tea caffeine by steep time.
How long caffeine stays in your body
Caffeine's half-life is the time your body takes to clear half of it. A 2017 review in Frontiers in Psychiatry gives 3–7 hours in adults and notes that serum levels peak about 2 hours after you drink it. The authors of a 2025 sleep trial used "typically 3–6 hours" when they set their washout period.
The liver does most of the work. A 2018 Pharmacological Reviews paper states that the CYP1A2 enzyme metabolizes 95% of the caffeine you take in, and that genetic variation in that enzyme is the main reason half-life differs so much between people.
The arithmetic is simple. Caffeine left = dose × 0.5^(hours ÷ half-life). We use 5 hours below because it sits in the middle of the 3–7 hour range. It is an assumption, not your number.
- 47 mg at 3 pm, 8 hours to an 11 pm bedtime: 47.4 × 0.5^(8 ÷ 5) = 15.6 mg, or 33% of the cup.
- The same mug at 5 pm leaves 21 mg; at 7 pm, 27 mg.
The formula ignores the roughly 2-hour absorption peak, so it slightly overstates what is left in the first hours and is close enough by bedtime.
Get caffeine and decaf notes
How much caffeine each tea type carries, how decaf tea is made, and caffeine-free alternatives. Free, no spam.
No fixed schedule. Unsubscribe anytime.
What changes your half-life
The 3–7 hour range is wide because several common conditions move it. These are the ones with published numbers.
| Factor | Measured effect | Source |
|---|---|---|
| Smoking | Doubles the rate of caffeine clearance | Temple 2017 |
| Smoking | Half-life 6.0 h in non-smokers vs 3.5 h in smokers (study cited in review) | Hereditas review 2026 |
| Quitting smoking | Clearance fell 36.1% within days in 12 heavy smokers | Faber & Fuhr 2004 |
| Oral contraceptives | Tend to double the half-life | Temple 2017 |
| Pregnancy | Half-life on average 8.3 h longer, up to 16 h longer | Temple 2017 |
| Late pregnancy | About 3 h in non-pregnant women vs up to 10.5 h in the final weeks | Hereditas review 2026 |
A 2021 systematic analysis of published caffeine pharmacokinetic data found the same direction across studies: smoking shortens half-life and oral contraceptive use lengthens it, over a wide range of doses. Liver disease, alcohol and some drugs also slow clearance, per the same reviews.
Here is what that spread does to one 47 mg mug drunk at 3 pm, measured at 11 pm:
| Half-life used | Caffeine left at 11 pm | Share of the cup |
|---|---|---|
| 3.5 h (smoker figure cited in the 2026 review) | 10 mg | 21% |
| 5 h (middle of the 3–7 h adult range) | 16 mg | 33% |
| 7 h (top of the adult range) | 21 mg | 45% |
| 10.5 h (final weeks of pregnancy, 2026 review) | 28 mg | 59% |
Sources: half-life values from Temple 2017 and the 2026 Hereditas review; per-cup caffeine from USDA FoodData Central. Remaining caffeine is our calculation.
What the sleep studies measured
Three studies give most of the timing evidence. Their doses are much higher than one cup of tea, which matters when you read the headlines.
| Study | Who | Dose and timing | Main result |
|---|---|---|---|
| Drake et al. 2013 | 12 adults (6 women, 6 men), aged 19–48 | 400 mg pill at 0, 3 or 6 h before bed, vs placebo | Total sleep time cut by 1.1–1.2 h at every time point, measured at home with a sleep monitor |
| Gardiner et al. 2023 | Meta-analysis of 24 studies | Various | Caffeine cut total sleep by 45 min and sleep efficiency by 7%; time to fall asleep up 9 min |
| Gardiner et al. 2025 | 23 men, mean age 25.3, habitual intake under 300 mg/day | 100 or 400 mg at 12, 8 or 4 h before bed, vs placebo | 100 mg: no significant effect at any time; 400 mg: delayed sleep onset even at 12 h |
Two details from these papers are easy to miss. In Drake's study, people taking caffeine 6 hours before bed did not report worse sleep in their diaries, but the monitor still recorded the loss. Gardiner 2025 found the same gap between felt and measured sleep and wrote that people "may have difficulty accurately perceiving the influence of caffeine on sleep quality."
The 2023 meta-analysis also modeled cut-off times. To avoid a drop in total sleep, it estimated that a 107 mg cup of coffee (250 mL) should be drunk at least 8.8 hours before bed, and a 217.5 mg pre-workout serving at least 13.2 hours before.
None of these studies tested tea. They tested caffeine pills or capsules. One 8 fl oz mug of black tea holds about 47 mg by USDA's figure, roughly a ninth of the 400 mg dose in Drake's study.
Caffeine left at bedtime, cup by cup
This table applies the 5-hour half-life to common drinks, for an 11 pm bedtime. Treat it as arithmetic, not a forecast of how you will sleep.
| Drink | Caffeine per serving | Left at 11 pm if drunk at 3 pm | at 5 pm | at 7 pm |
|---|---|---|---|---|
| Decaf black tea, 8 fl oz | 2 mg | <1 mg | 1 mg | 1 mg |
| Green tea, 8 fl oz | 28 mg | 9 mg | 12 mg | 16 mg |
| Oolong tea, 8 fl oz | 38 mg | 13 mg | 17 mg | 22 mg |
| Black tea, 8 fl oz | 47 mg | 16 mg | 21 mg | 27 mg |
| Matcha, 2 g (high end of range) | 89 mg | 29 mg | 39 mg | 51 mg |
| Brewed coffee, 8 fl oz | 95 mg | 31 mg | 41 mg | 54 mg |
Sources: per-serving values from USDA FoodData Central (8 fl oz = 237 g) and the matcha range of 18.9–44.4 mg/g in our caffeine comparison. Remaining caffeine is our calculation at a 5 h half-life.
For scale, the 2025 trial's 100 mg dose taken 4 hours before bed would leave about 57 mg at bedtime under the same assumption, and that dose showed no significant sleep effect in those 23 men. A 7 pm mug of black tea leaves about 27 mg. Two mugs at 7 pm leave about 54 mg. A cup of coffee at 3 pm leaves about 31 mg.
Steep time moves these numbers. A 5-minute steep extracts more caffeine than a 2-minute one; our steep-time article has the extraction curves.
Evening options: decaf, herbal, and low-caffeine tea
Three ways to drink tea late without much caffeine:
- Decaf tea. USDA lists decaffeinated brewed black tea at about 2 mg per 8 fl oz. Decaf is not caffeine-free; methods and residues are in how tea is decaffeinated.
- Herbal infusions. Chamomile, rooibos and honeybush are not made from the tea plant (Camellia sinensis). See rooibos and honeybush.
- Smaller, shorter cups of real tea. Less leaf and a shorter steep both reduce caffeine. Applying the formula above, halving the dose halves what is left at bedtime.
L-theanine, the amino acid in tea that is often sold for sleep, has its own trial record. We covered the doses and the industry funding behind those trials in L-theanine and caffeine: what the studies show, and a cup of tea contains far less than the trial doses.
Chamomile: what the trials found
The evidence for chamomile and sleep is mixed. The reviews disagree because they pool different people, different products and different outcomes.
- Kazemi et al. 2024: 10 studies, 772 participants. Pooled Pittsburgh Sleep Quality Index (PSQI) score fell by 1.88 points across 5 studies, with high heterogeneity (I² 88.4%). Sleep duration, sleep efficiency and daytime functioning did not improve. Only one study checked whether blinding worked.
- Hieu et al. 2019: 12 randomized trials. Sleep quality improved (standardized mean difference −0.73). Only one trial measured insomnia severity, and it found no significant change.
- Leach & Page 2015: 14 trials, 1,602 people with insomnia, covering valerian, chamomile, kava and wuling. No herbal medicine beat placebo on any of 13 efficacy measures.
- Chang & Chen 2016: one of the few trials that used chamomile tea itself. 80 Taiwanese postnatal women with poor sleep, 40 drinking chamomile tea for 2 weeks and 40 on usual care (no placebo drink). Sleep-related scores improved at 2 weeks, but the groups were similar again at 4 weeks.
What the reviews support: chamomile may improve self-rated sleep quality a little in some groups. What they do not support: longer sleep, or a treatment for insomnia. A cup of chamomile does carry one plain advantage in the evening: it is not a caffeine source.
Frequently Asked Questions
How long does the caffeine in a cup of tea last?
Caffeine's half-life in adults is 3–7 hours, per a 2017 review. At 5 hours, an 8 fl oz mug of black tea with about 47 mg leaves about 16 mg 8 hours later. Smoking shortens the half-life; oral contraceptives and pregnancy lengthen it.
What time should I stop drinking tea before bed?
No study has tested tea itself. The trials used caffeine pills: 400 mg cut sleep by more than an hour even 6 hours before bed in 12 adults, while 100 mg taken 4 hours before bed showed no significant effect in 23 young men. A mug of black tea holds about 47 mg, so most of the timing advice in the press is built on doses 2 to 9 times larger.
Is green tea better than black tea at night?
USDA's per-cup figures are about 28 mg for green tea and 47 mg for black tea. At a 5-hour half-life, a 7 pm cup leaves about 16 mg and 27 mg respectively at 11 pm. Brew strength can erase that gap, so the steep matters as much as the type.
Does chamomile tea help you sleep?
Reviews disagree. A 2024 meta-analysis of 10 studies (772 people) found a 1.88-point improvement on the PSQI sleep-quality score but no change in sleep duration, and a 2015 review of 14 insomnia trials found no herbal medicine better than placebo. The one 80-person trial of chamomile tea found benefits at 2 weeks that were gone by 4 weeks.
Is decaf tea caffeine-free?
No. USDA lists decaffeinated black tea at about 2 mg per 8 fl oz, against about 47 mg for regular black tea. At a 5-hour half-life, a 7 pm cup of decaf leaves under 2 mg at 11 pm.
