Educational article. Not medical advice. We make a sleep formula with no melatonin in it — our conflict of interest is spelled out at the end. Every trial cited here carries its PubMed ID so you can read it yourself.
There is a sentence we read more often than any other in the sleep groups we follow, and it is almost always phrased the same way: "Melatonin worked for the first couple of weeks, then it just stopped." It comes with a ritual attached — the switch to a higher dose, the switch to a gummy, the switch to a "time-release" one, the week off to "reset," the return, the same result. And it produces a very reasonable question, which is the one we're going to answer: if melatonin stops working, is there anything that does the opposite?
The honest answer has two halves, and the first one is inconvenient for us.
Half one: the trials say melatonin doesn't fade
We went looking for the tolerance everyone reports, and it isn't there. The largest meta-analysis of melatonin for sleep — 19 randomized trials, 1,683 people — puts the effect at about 7 minutes faster to fall asleep and about 8 minutes more sleep per night, with a small improvement in rated sleep quality. Its own conclusion, word for word: the effects "are modest but do not appear to dissipate with continued melatonin use" (Ferracioli-Oda et al., PLoS One 2013, PMID 23691095).
A six-month trial of prolonged-release melatonin in 555 adults found the same thing from the other direction: improvements were "maintained or enhanced over the 6-month period with no signs of tolerance," and no rebound when people stopped (Wade et al., Curr Med Res Opin 2011, PMID 21091391).
The closest thing to a mechanism for "it stopped" lives in a dish, not a person. Melatonin receptors in cultured cells do down-regulate after exposure, and after a high, supplement-like concentration they recover more slowly — not fully back after 24 hours (Gerdin et al., FASEB J 2004, PMID 15522910). That is interesting, and it is a real reason to be suspicious of 10 mg gummies. It is not evidence that the clinical effect disappears, because the clinical trials measured exactly that and it didn't.
So we're not going to tell you melatonin wears off. We'd be making it up.
Half two: it never grew, either
Look at those numbers again. Seven minutes. Eight minutes. That is the entire effect, and — this is the part that matters — it is the same effect on night one and on night ninety. Melatonin is a timing signal. It tells the clock in your brain that night has started. It has nothing to say to a stress system that idles back up at 3 a.m., which is why the American Academy of Sleep Medicine's guideline suggests clinicians not use it for staying asleep, and not for falling asleep either (Sateia et al., J Clin Sleep Med 2017, PMID 27998379).
Which means the memory "it worked, then it stopped" is real, but it's mislabeled. Here is what the trials suggest actually happened. The first nights on anything new carry a large, well-documented expectation effect: in sleep drug trials the placebo alone measurably shortens time to fall asleep and lengthens sleep (Winkler & Rief, Sleep 2015, PMID 25515108; Yeung et al., Sleep Med Rev 2018, PMID 28554719). That part wears off. The seven minutes never did. If your problem was 3 a.m., melatonin didn't stop doing the job. It never started. What ended was the honeymoon on a tool that was aimed at a different problem.
That reframe is the whole article, really. Melatonin doesn't fade. It just never grows. And "does it grow?" turns out to be the question that separates sleep ingredients better than almost any other.
What a curve that grows looks like
Three of the best-evidenced non-melatonin sleep ingredients share a property melatonin doesn't have: in their trials, the gap from placebo was widest late.
Ashwagandha root extract, 600 mg a day. In a 10-week trial with wrist actigraphy, time to fall asleep improved in both the ashwagandha and placebo groups at week 5 — and the difference between them only reached significance at week 10, when sleep efficiency in the ashwagandha group had climbed from 75.6% to 83.5% (Langade et al., Cureus 2019, PMID 31728244). An 8-week trial in healthy adults and poor sleepers found the same direction (Langade et al., J Ethnopharmacol 2021, PMID 32818573). The meta-analysis across trials says the effect is more prominent at doses of 600 mg a day or more and durations of eight weeks or more (Cheah et al., PLoS One 2021, PMID 34559859). The likely reason it takes that long: the same extract, at the same dose, lowered serum cortisol over 60 days in adults under chronic stress (Chandrasekhar et al., Indian J Psychol Med 2012, PMID 23439798) — and the stress axis is a slow system to talk down.†
Saffron extract, 28 mg a day. The trial took its readings at days 7, 14, 21 and 28, and the improvement in rated sleep quality over placebo accrued across that month, not on the first night (Lopresti et al., J Clin Sleep Med 2020, PMID 32056539).†
L-theanine, 200 mg a day. Measured after four weeks: better scores on the Pittsburgh Sleep Quality Index in a crossover trial of 30 adults (Hidese et al., Nutrients 2019, PMID 31623400). Theanine also has an acute, same-evening calming effect in other work, which is why we call it an "early mover" — but the sleep-quality reading that counts was taken at week four.†
The practical consequence: judge night 30, not night 1
Melatonin trained a generation of people to evaluate a sleep aid on the first night. For a timing signal, that's fair: what you get on night one is what you get. For ingredients whose trials measured their biggest gap at weeks eight to ten, the first-night test is exactly backwards. The honest expectation runs like this:
- Nights 1–7: mostly quiet. Theanine and saffron are the early movers — a calmer wind-down, not a knockout.†
- Nights 8–28: "wired at midnight" has less to work with. This is the whole window of the saffron study, and theanine's.†
- Weeks 5–10: where the ashwagandha trials found their biggest gap from placebo.†
- Night 90: the end of the studied range. Beyond it nobody has data — including us.
If you track your sleep with a ring or a wristband, this changes what to look at: the seven-day trend of time awake, not any one night. One bad night on night 4 proves nothing in either direction; a flat trend at night 30 is real information.
What we don't know
Two gaps, stated plainly. First, none of these trials followed people after they stopped. Whether the change holds without the capsules, fades slowly, or fades fast is an open question, and anyone who tells you otherwise is guessing. Second, none of them ran past ten weeks. "Keep taking it and it keeps getting better" is an extrapolation past the last data point; the studied window is roughly ninety nights, and that's why our protocol is ninety nights rather than forever.
And one more, because it's the honest version of the melatonin story: the placebo effect that made melatonin's first week feel good will make any new capsule's first week feel good too, including ours. The difference isn't that we're immune to it. The difference is what the curve does after it wears off.
Disclosure, as every time: we make Night Reset, a melatonin-free evening formula built around these three ingredients at the doses above — 600 mg KSM-66® ashwagandha, 200 mg L-theanine, 28 mg affron® saffron. An article arguing that melatonin's effect is flat and that these ingredients build over weeks obviously flatters our product. That is exactly why every claim above carries a PubMed ID, and why the section you just read is titled "What we don't know."
†These statements have not been evaluated by the Food and Drug Administration. This product is not intended to diagnose, treat, cure, or prevent any disease. This article is general education, not medical advice — decisions about supplements and medication belong with you and your doctor.