Low physiological toxicity is the part everybody quotes. The interaction risks and the psychiatric exclusions are the part that decides whether any of this is safe for a particular person, and they are specific enough to be listed.
The short answer
Psilocybin has low physiological toxicity but real psychiatric and interaction risks, and those risks are concentrated in specific, identifiable groups. Lithium is a genuine contraindication associated with seizures.[1] MAOIs are excluded in practice. A personal or family history of psychosis or bipolar I is an exclusion in essentially every trial, which also means the published safety data do not describe that group. For everyone else the acute adverse effects are mostly transient, and the best summary of them is a 2024 systematic review and meta-analysis.[2]
Who this is not safe for
Each row carries the strength of the evidence behind it, because a warning drawn from a survey and a warning drawn from a trial protocol are not the same kind of claim and should not read as though they were.
Naturalistic use of psilocybin alongside lithium is associated with seizures. The underlying study is survey-based, which is a real limitation — but the signal is strong and the outcome is severe, and no responsible program accepts this combination.[1]
Monoamine oxidase inhibitors are an exclusion in trial protocols and in the legal state programs. This is a pharmacological interaction of a serotonergic compound with a drug class that blocks its metabolism, and it is not something to work out by experiment.
This is the single most consistent exclusion criterion in the psilocybin literature. It means two things at once: there is a plausible risk of precipitating an episode, and the published safety data therefore do not describe people with these histories at all. Nothing in the trials tells you what happens in this group, because the trials deliberately did not enroll them.
Psilocybin typically raises blood pressure and heart rate, and the National Institute on Drug Abuse states that this may be dangerous for people with heart conditions. In healthy volunteers the rise is transient and unremarkable. The risk is a pre-existing condition that makes a transient rise consequential, which is the reason low physiological toxicity should never be read as no physiological effect.[11]
Antipsychotics act at the receptors psilocin acts on and will blunt or block the experience. The practical hazard is a person taking more in response to nothing happening, or stopping a medication in order to make it work.
A 2024 analysis found psilocybin's advantage held in never-medicated participants but disappeared among people who had recently discontinued an SSRI. It inverts the common advice to taper quickly before a session. It is post-hoc, so treat it as a caution rather than a finding.[7]
This list is not a screening tool and is not exhaustive. It is the set of exclusions that appear consistently across trial protocols and the legal state programs, with the evidence behind each one stated. Screening is done by a clinician who takes a history.
The evidence on comparative harm
A UK expert panel scored twenty drugs on sixteen measures, covering harm to the person taking them and harm to everyone around them. Mushrooms came last. This chart is used to sell psychedelic services more than any other single figure, so it is worth being precise about what it does and does not establish.
Overall harm score, 0–100, across 16 criteria — 9 for harm to the user and 7 for harm to other people. Eight of the 20 substances scored, shown in the paper’s rank order. Nutt DJ, King LA, Phillips LD. Drug harms in the UK: a multicriteria decision analysis. The Lancet 2010;376:1558–65.
Read from the paper’s figure rather than quoted from its text. Alcohol, heroin and crack cocaine are stated numerically in the abstract; the rest are graphed, and reproductions differ by a point — psilocybin mushrooms appears as both 6 and 5. The rank order is not in dispute.
The first four of these are the paper’s own stated limitations, quoted. We are not undermining the study — it is a careful piece of work that says plainly what it cannot do. We are reprinting the part that gets cropped out.
The authors are explicit: “We scored only harms. All drugs have some benefits to the user, at least initially, otherwise they would not be used.” A low harm score is not a high benefit score, and this study makes no claim at all about whether anything on it works.
“Many of the harms of drugs are affected by their availability and legal status.” Part of what a prohibited drug scores is the harm of prohibiting it — unregulated supply, adulteration, criminal exposure. The paper says a model ideally ought to separate harm from the drug and harm from the control system, and that this one does not.
The results “do not relate to drugs when used for prescription purposes.” The study is about drugs in a population, obtained and taken in the ordinary way. It is not a comparison of medicines.
The authors note the results “are not necessarily applicable to countries with very different legal and cultural attitudes to drugs.” It is a structured expert judgment exercise from 2010 — a good method, honestly reported, and still a judgment rather than a measurement.
A score averaged across everyone who uses a substance cannot tell you what happens to a particular person with a family history of bipolar I, or taking lithium, or with an unscreened cardiac problem. The risks that decide whether psilocybin is safe for you are concentrated in identifiable groups, and they are not visible anywhere on this chart.
The analysis was published in 2010. Mephedrone was a headline drug in it. The psilocybin evidence base that this site reports on barely existed. Nothing about the ranking has been overturned, but a chart this old being the strongest safety claim in a category's advertising is itself worth noticing.
The best current summary is a systematic review and meta-analysis of acute adverse effects at therapeutic doses, published in JAMA Network Open in 2024.[2] That design matters: it pools results across trials rather than reporting one, which is the right way to read adverse-event data, because any single trial is too small to characterize anything but the common effects.
What it covers is the acute period — the session itself and the hours around it — in people who passed trial screening. Those are two significant boundaries. It does not tell you about the groups the trials excluded, and it does not tell you about anything that emerges months later. We are not going to quote a percentage out of it here, because a pooled figure detached from its inclusion criteria is exactly the kind of number that ends up in a headline meaning something it does not mean. Read the source.
The other honest data point is real-world rather than experimental. In Oregon’s regulated program, 2% of clients described the experience as harmful at three months.[3] There is no control group, and participants were healthier than typical trial populations, so that figure describes what happened rather than what psilocybin caused. It is still the largest look anyone has at a legal program operating at scale.
The federal summary of the same ground is worth reading beside both. The National Institute on Drug Abuse states that psilocybin typically raises blood pressure and heart rate, which may be dangerous for people with heart conditions, and that agitation, confusion, nausea and vomiting can be severe enough to require medical attention.[11] It also treats what people call a bad trip — extreme fear, anxiety, panic or paranoia — as a risk in its own right rather than as an inconvenience. None of that is incompatible with low physiological toxicity. They are answers to different questions.
The same page locates the risk where the rest of this site keeps finding it. NIDA puts the risk of psychosis or suicidality low in supervised clinical settings, and places the health risks in use that is unsupervised and outside a research or clinical setting.[11] Two of those risks are not pharmacological at all: misidentifying a mushroom and taking a toxic one instead, and commercial products marketed as containing psilocybin that have been found to contain toxic chemicals. Neither is a risk a supervised program carries, and neither is addressed by anything in the trial literature.
Galerina marginata, the funeral bell. It contains the same amatoxins as the death cap, it grows on wood in the same months and the same places as several Psilocybe species, and at a glance it looks like them. This is the single most important photograph on this page.
This is one of the few questions in the field with a straightforward answer from a government source. NIDA states that research to date suggests use of psilocybin does not typically lead to addiction, and that DSM-5 does not include substance use criteria specifically related to psilocybin, though it does include a diagnosis of other hallucinogen use disorder.[11]
Two qualifications belong next to it. The absence of a psilocybin-specific diagnostic category is partly a fact about how the manual was written rather than a measurement of the drug, and the same page reports no withdrawal data in either direction. Low addictive potential is also the least consequential thing on this page. The risks that decide whether psilocybin is safe for a particular person are concentrated in a single session, and they do not shrink because there is no habit at the end of it.
The blinding failure in psychedelic research is usually discussed as a threat to efficacy estimates. Participants can tell whether they received a psychedelic, so can the people rating them, and expectation inflates the apparent benefit. The VA names this explicitly in its own assessment.[4]
The same mechanism runs in the other direction on harms. If everyone in the room knows who got the drug, adverse events in the psilocybin arm can be attributed to the drug and adverse events in the placebo arm attributed to something else — or the reverse, if the investigators are invested in the compound. Unblinded safety reporting is soft in both directions, and the softness is not quantified.
This is why the alcohol use disorder trial matters beyond its own result. It used an active placebo that produces noticeable effects of its own, which is the only structural fix anyone in this field has actually implemented — and its authors still reported that the blind was not fully maintained.[5]
This is the question we are asked most, and the one where bad advice does the most damage.
On interaction. In a controlled study, two weeks of escitalopram pretreatment reduced anxiety and bad drug effects while leaving positive mood unchanged, and no serotonin syndrome was observed.[6] That is reassuring as far as it goes. It is one study, in a pretreatment design, and it does not license anyone to combine these at home.
On discontinuation. The widespread advice is to taper off an antidepressant before a session so the psilocybin “works properly”. A 2024 post-hoc analysis inverts it: psilocybin’s advantage over escitalopram held in never-medicated participants but disappeared among people who had recently discontinued.[7] Post-hoc analyzes are hypothesis-generating rather than conclusive. But the direction of the finding is the opposite of the advice being given, which is reason enough to stop repeating the advice.
On the cost of stopping. The ANTLER trial reported 56% relapse within a year after discontinuation against 39% on maintenance.[8] Withdrawal effects are also routinely mistaken for the illness returning, which makes the experience of stopping harder to interpret than people expect. Discontinuing a medication to qualify for a psilocybin session is a decision with a measurable downside, taken to access something with a smaller measured upside than the marketing implies.
Every legal pathway in the United States is built around a supervised administration session lasting roughly six to eight hours, with a trained facilitator present throughout.[9] That requirement is not ceremony and it is not therapy theater. It is the safety model.
For that period a person cannot reliably assess risk, leave, or ask for help. The acute effects that most often go wrong — anxiety escalating into panic, attempts to leave a safe environment — are containable by a person who is present and are not containable otherwise. Federal guidance to states says plainly that prescribing psilocybin for unsupervised home use is not supported by the current evidence.[10]
The corollary is uncomfortable and worth stating: the supervision requirement is most of what makes this expensive. A full day of a trained person’s time is the dominant cost in every legal program, which is set out on our cost page. Anyone offering the same thing much cheaper is removing the part that makes it safe.
We do not publish doses. We do not publish taper schedules. We do not publish sourcing advice, cultivation instructions, or how to obtain psilocybin where it is illegal. We will not make an exception for a sympathetic case, and we do not treat this as a limitation of the site.
The reasoning is simple. Each of those is a decision that depends on an individual history — what else you take, what runs in your family, what you have been diagnosed with — and a page cannot know any of it. Publishing a number that looks like guidance would substitute our judgment for a clinician’s at exactly the point where the clinician’s is the only one worth having. This is written down in full on our editorial standards page, so that it is a rule rather than a mood.
What we will do is report what the trials found, who they excluded, and how strong the evidence is — which is on the research page — and tell you where any of this is legal, on the state tracker.
Common questions