Every major trial, with the things that decide whether a result means anything — design, sample size, comparator, and whether the primary endpoint was met — reported next to the headline rather than after it.
The short answer
The evidence for psilocybin is real, modest, and consistently overstated — and the strongest trial in the field is in alcohol use disorder, not depression. Two Phase 3 depression trials met their primary endpoints with differences of 3.6 and 3.8 MADRS points, which is genuine and about the size of a conventional antidepressant effect. The field’s most-quoted trial did not meet its primary endpoint at all. There is no published trial data on psilocybin for PTSD, a condition it is nonetheless widely marketed for.
By condition
Ranked by the quality of the underlying trials rather than by how much attention each one gets. These two orders are almost inverted.
The best-designed trial in the field is in alcohol use disorder, not depression: 95 participants, double-blind, with an active placebo rather than an inert one, showing heavy drinking days of 9.7% against 23.6% over 32 weeks.
Our read: The only trial to use an active placebo, which is the field's central methodological weakness everywhere else.
Two Phase 3 trials of synthetic psilocybin met their primary endpoints in 2025 and 2026, with modest differences of 3.6 and 3.8 MADRS points. The most-quoted trial in the field — psilocybin versus escitalopram — did not meet its primary endpoint.
Our read: The largest evidence base, and the most misreported. Effect sizes are real, statistically significant, and modest.
Psilocin acts primarily at the serotonin 5-HT2A receptor, and imaging studies report changes in brain connectivity and structure that outlast the acute effects. The mechanism by which any of this would produce durable clinical benefit is not established.
Our read: Small, exploratory, frequently unreplicated studies. Mechanistic plausibility is not evidence of benefit.
There is no published clinical trial data on psilocybin for PTSD. The US Department of Veterans Affairs states this plainly, and neither psilocybin nor MDMA appears in the VA/DoD Clinical Practice Guideline for PTSD.
Our read: Widely marketed for PTSD anyway. This is the clearest gap between what is sold and what is known.
The strength rating is our editorial judgment, not a formal GRADE assessment, and we label it that way wherever it appears. The underlying trial characteristics in the table below are facts; the ranking is a reading of them.
Find your condition
Ten grouped categories across every psilocybin trial currently enrolling. Each one shows our assessment where we have published it, and says so where we have not.
What the evidence shows
Promising, modest effects
The largest evidence base, and the most misreported. Effect sizes are real, statistically significant, and modest.
Depression in fullPhase 3 · VA Office of Research and Development
Phase 3 · Sharmin Ghaznavi · Requires a medication washout
Phase 2 · Alan Davis
Phase 2 · Charles S. Grob, M.D.
The ledger
Sample size, control group, blinding and primary endpoint sit beside the result rather than in a paragraph underneath it. Read the two together.
| Trial | Design | Participants | Primary endpoint | Result and limitation |
|---|---|---|---|---|
| COMP005 and COMP006 (Compass Phase 3)2026 · Treatment-resistant depressionSource ↗ | Two randomized Phase 3 trials of synthetic psilocybin (COMP360) ControlledDouble-blind Compared against: Placebo (COMP005); 1 mg active comparator (COMP006) | —Not published in full | Met | Both met their primary endpoint. The differences were −3.6 MADRS points versus placebo at week six, and −3.8 points for 25 mg against 1 mg. Read it with this: Topline results released by the sponsor, not yet published in full. The effect sizes are modest and in the same broad range as conventional antidepressants in acute trials. |
| Psilocybin for alcohol use disorder2022 · Alcohol use disorderSource ↗ | Randomized, double-blind trial with 12 psychotherapy sessions in both arms ControlledDouble-blind Compared against: Active placebo (diphenhydramine) | 95 | Met | Heavy drinking days 9.7% versus 23.6% over 32 weeks — a mean difference of 13.9 percentage points (95% CI 3.0–24.7, p=0.01, Hedges g=0.52). Abstinence 47.9% versus 24.4%. Read it with this: The blind was still not fully maintained, and the psilocybin effect cannot be separated from the twelve psychotherapy sessions both arms received. |
| Single-dose psilocybin for major depressive disorder2023 · Major depressionSource ↗ | Randomized trial of a single 25 mg dose ControlledDouble-blind Compared against: Niacin | 104 | Met | A significant reduction in depression scores against niacin over six weeks. Read it with this: Participants had to discontinue antidepressants for at least two weeks or five half-lives before dosing, so the result does not describe people currently taking medication. |
| Psilocybin versus escitalopram2021 · Major depressionSource ↗ | Randomized head-to-head against a standard antidepressant ControlledDouble-blind Compared against: Escitalopram | 59 | Not met | The primary outcome was not statistically significant: a difference of −2.0 points (95% CI −5.0 to 0.9, p=0.17). Read it with this: The famous response and remission figures from this trial are unadjusted secondary outcomes. The authors state explicitly that no clinical conclusions can be drawn from them. |
| Psilocybin alongside a continuing SSRI2023 · Treatment-resistant depressionSource ↗ | Open-label study in people who stayed on their SSRI No control groupOpen-label Compared against: None | 19 | No primary endpoint | Response was observed in participants who did not discontinue their antidepressant. Read it with this: Nineteen people, open-label, no control group. It shows SSRIs do not obviously abolish a response. It does not show they improve one, and it cannot estimate effect size. |
| Oregon state-regulated psilocybin services2026 · Major depressionSource ↗ | Observational study of a real-world legal program No control groupNot blinded Compared against: None | 346 | Not applicable | 91.5% of clients reported benefit at one month; 2% described the experience as harmful at three months. Read it with this: No control group, so it cannot attribute benefit to psilocybin. Participants were healthier than typical trial populations, with only about half reporting moderate-to-severe depression at baseline. |
| Brain changes after first psilocybin use2026 · Healthy volunteersSource ↗ | Placebo-controlled imaging study, fixed order ControlledSingle-blind Compared against: Placebo | 28 | No primary endpoint | Measurable changes in brain structure persisting beyond the acute drug effect. Read it with this: Twenty-eight healthy volunteers, exploratory, not pre-registered, and fixed-order rather than randomized. The authors explicitly call for replication. |
The ledger lists trials individually rather than pooling them. For depression, the Metapsy project maintains a living meta-analytic database of these randomized trials that will pool them on request, with heterogeneity, publication-bias and risk-of-bias output alongside.[5] Its figures are computed when a reader runs an analysis rather than stated on the page, so we quote none of them here — but a pooled estimate from a database of that kind outranks any single trial in the table above, and it is the right place to check anything we say against.
Uncertainty
A trial reports one number, but the real finding is a range. Where that range sits tells you whether a result survives its own uncertainty. Only 2 of the 7 trials above publish one; the rest are listed below with the reason.
This is a real difference. The whole range of possible answers sits on one side of zero, so the effect holds up even at the pessimistic end. Scale: percentage points of heavy drinking days. Below zero favors fewer heavy drinking days on psilocybin. p=0.01. Hedges g=0.52.
This might be nothing. The range of possible answers still includes zero, so these data are also compatible with the treatment doing nothing at all. Scale: QIDS-SR-16 points versus escitalopram. Below zero favors lower depression score on psilocybin. p=0.17.
Why the other 5 are absent
That distribution is the honest state of this literature. Filling the gaps with figures converted between scales, or read off a published graph, would manufacture the precision this page exists to question.
Almost every criticism worth making of this literature reduces to one of two things, and neither is a reason to dismiss the research. They are reasons to size the claims correctly.
Blinding does not hold. Participants can tell whether they received a psychedelic — the effects are unmistakable — and so can the people rating them. That inflates apparent benefit through expectation in ways trials rarely quantify. Most trials compare against an inert placebo, which makes the problem worse rather than better. The alcohol use disorder trial is the field’s honorable exception: it used diphenhydramine, an active placebo that produces noticeable effects of its own, and the authors still reported that the blind was not fully maintained.[1]
The results describe unmedicated people. Most trials required participants to be off antidepressants before dosing.[2] Since roughly one in eight American adults takes one, the published effect sizes do not describe a large share of the people reading about them. This is not a footnote. It is a limit on who the numbers apply to.
Psilocybe azurescens on the Oregon coast. Everything on this page was measured in a synthesised, precisely weighed compound given under supervision — not in this. The mushroom is where the molecule was found, not what the trials studied.
Stating this in advance is the only way a judgment like ours can be held to account later. We would revise upward on: a large trial using an active placebo that replicates the depression results; published data in PTSD of any quality; or durable outcomes beyond twelve months in a controlled sample. We would revise downward on: failure to replicate the Phase 3 results, or a serious adverse-event signal emerging from Oregon or Colorado at scale.
Common questions