The grades below cover acute use, meaning effects in the hours after a dose. Our evidence grading method explains what each grade and maturity label requires.
Does caffeine improve attention?
Yes, in the short term. A 2025 meta-analysis pooled 31 randomized, double-blind, placebo-controlled trials with 1,455 rested healthy adults and found caffeine improved accuracy on attention tests (g = 0.27, a small effect) [2]. The effect was not related to how much caffeine people usually consumed, how complex the task was or which attention network was tested [2]. That review was publicly funded by Danmarks Frie Forskningsfond, and its authors declared no conflicts [2].
Dose matters, but not in a straight line. Doses of 200 mg or more improved accuracy more than lower doses, yet accuracy improved only up to a point before declining [2]. A military and occupational review found alertness, vigilance and attention improve after low (about 40 mg) to moderate (about 300 mg) doses, while effects on memory and judgment are less consistent [1].
Other populations show the same direction. A meta-analysis of 45 publications in sleep-deprived people found better accuracy on attention tests (g = 0.68) [3], and a Cochrane review of 13 mostly simulated trials in shift workers found better orientation and attention (SMD -0.55), with a high risk of bias [19].
One preregistered trial (NCT02461693, 205 adults, a single 200 mg dose against placebo) was designed to separate the drug's effect from expectation [52]; its authors reported that some effects differed by whether people expected the active drug [20].
Does caffeine improve reaction time?
In trials, yes. In the same 31-trial meta-analysis, caffeine shortened reaction time on attention tests (g = 0.28), and unlike accuracy, reaction time kept improving as the dose rose [2]. The military and occupational review also found faster reaction time after low to moderate doses [1].
Does caffeine improve endurance?
Yes, for many kinds of exercise. An umbrella review of 21 meta-analyses found moderate-quality evidence that caffeine improves aerobic endurance, muscle endurance, strength and anaerobic power, with larger effects for aerobic exercise [5]. The corrected version of a 2018 meta-analysis of 44 placebo-controlled studies reports an overall effect size of 0.28 for time-trial completion time and a small effect at moderate doses of 3-6 mg/kg [6].
Newer pooled studies agree on direction but not fully on the lowest helpful dose. A 2024 meta-analysis of 15 crossover studies found 4-6 mg/kg cut completion time (SMD -0.55) while 1-3 mg/kg did not reach significance [21]. Later analyses of 20 studies [23], of 48 studies with 689 participants [32] and of 48 studies in a network meta-analysis (SMD -0.34 for low-dose capsules) [33] found that 3 mg/kg or less did cut completion time, though one found moderate doses more consistent [32]. A 2026 meta-analysis of 23 double-blind trials in cyclists found a small benefit (SMD -0.32, I² = 0%) [34].
Most participants were young men [5]; a 2026 meta-analysis of 19 crossover trials in 235 women found a small benefit (SMD 0.21), with risk of bias rated "some concerns" [22]. Most trials are small crossovers, but one double-blind trial randomized 101 competitive male athletes and was registered before it began, with an industry collaborator [31][54]. Overall, 4 mg/kg cut 10-km cycling time by 3%, but in men with the CC genotype of the CYP1A2 gene it increased cycling time by 13.7%, and the AC group showed no effect [31]. The 2018 review also found two studies with slower times [6]. The evidence is mixed by individual.
Does caffeine delay sleep onset?
Yes. This grade describes the strength of evidence for a harmful effect. In studies of night-time sleep after doses taken within hours of bed, a 2023 meta-analysis of 24 studies found caffeine added 9 minutes to the time taken to fall asleep and 12 minutes awake after falling asleep [8]. A 2025 meta-analysis of 22 crossover trials of acute doses (956 healthy adults) found 8.35 minutes [10]. High and low doses both lengthened sleep onset, and differences by age and dose were not statistically significant [10].
Timing matters. In a placebo-controlled home study, 400 mg taken at bedtime, 3 hours or 6 hours before bed all disturbed sleep [9]. The 2023 meta-analysis estimated that total sleep was not reduced when a 107 mg coffee was finished at least 8.8 hours before bed, or a 217.5 mg pre-workout serving at least 13.2 hours before [8].
Does caffeine shorten total sleep?
Probably, when taken within hours of bedtime. This is also an adverse grade. Caffeine cut total sleep time by 45 minutes, sleep efficiency by 7% and deep sleep by 11.4 minutes in the 2023 meta-analysis [8], and total sleep by 34.67 minutes, in young and middle-aged groups, in the 2025 one [10].
Not every study agrees. A 2025 meta-analysis of 10 studies in 128 athletes taking caffeine before late-afternoon or evening training found a non-significant trend toward less total sleep (-32.47 minutes, p = 0.08) [35]. In a trial of 20 male habitual users taking 150 mg three times a day for 10 days, total sleep did not differ from placebo; recording began 5 hours after usual bedtime [36].
Does tolerance cancel the benefit?
Partly. In a crossover trial of 11 active adults taking 3 mg/kg daily for 20 days, caffeine raised peak cycling power by about 4% for the first 15 days, then the effect lessened [24]. In a crossover trial of 12 volunteers given 200 mg twice daily for a week, tolerance was incomplete, and cognitive performance was not affected by previous caffeine exposure [25].
For mental performance, a focused review argues that placebo-controlled studies cannot separate a true benefit from reversal of withdrawal in regular users, and that long-term withdrawal studies find little net benefit over abstinence [4]. Against that, the 2025 meta-analysis found the attention benefit did not vary with habitual intake [2]. Our grades cover acute effects only.
What happens when regular users stop?
A review of 57 experimental and 9 survey studies confirmed withdrawal symptoms including headache, fatigue, low energy, drowsiness, low mood, difficulty concentrating and irritability [13]. Symptoms typically began 12-24 hours after the last dose, peaked at 20-51 hours and lasted 2-9 days, and headache occurred in 50% of people in experimental studies [13]. Stopping intakes as low as 100 mg a day produced symptoms [13].
Caffeine plus L-theanine
Tea combines caffeine with the amino acid L-theanine. A 2014 meta-analysis of 11 placebo-controlled studies found moderate effects favoring the pair on alertness and attention-switching accuracy in the first 2 hours [26]. A 2025 meta-analysis co-funded by Lipton Teas and Infusions, with authors employed by Lipton and Unilever, found better attention-switching accuracy in the second hour (SMD 0.33), with wide uncertainty around most estimates [27]. The pair is graded on our L-theanine page and covered in our guide to caffeine and L-theanine.
What the evidence does not show
Effects on memory and complex judgment are less consistent than effects on alertness and reaction time [1]. Whether daily, long-term use gives a net mental benefit over not using caffeine at all remains disputed [4].
How sure are we?
On November 15, 2026 we searched all completed caffeine records in ClinicalTrials.gov (261 exercise and 122 sleep records), all 302 in ISRCTN, and PubMed for registered trials. Each preregistered endurance trial we found with published results had an industry funder, collaborator or product donor [31][54], and we found none for sleep in healthy adults, so total sleep time is Emerging [10]. Science is provisional. These grades reflect the evidence we found as of November 15, 2026.