Psychedelic safety discourse has two favorite settings: “basically harmless” and “obviously dangerous.”
Reality, as usual, has declined the invitation to become a slogan.
A study published September 29, 2026 in JAMA Network Open examined self-reported adverse health outcomes among U.S. adults who had recently used psilocybin, LSD, MDMA, ibogaine or 5-MeO-DMT. It is useful because controlled clinical trials do not tell us everything about what happens when these substances are used in the general population.
It is also exactly the kind of study that can be mangled by a headline if the limitations disappear.
What researchers studied
The analysis used data from the 2025 National Survey Investigating Hallucinogenic Trends, a double opt-in online panel survey of adults across all 50 states. The data were weighted against a nationally representative anchor survey.
Among 1,964 adults in the analysis, reported past-year use included 1,151 respondents for psilocybin, 376 for LSD, 539 for MDMA, 74 for ibogaine and 95 for 5-MeO-DMT. People who reported recent use were asked about physical and mental-health changes, health-care use and a set of symptoms they attributed to each substance.
The study found that self-attributed adverse outcomes were most frequent among respondents reporting ibogaine or 5-MeO-DMT use. Because psilocybin, LSD and MDMA were more commonly reported, the estimated total number of people experiencing adverse outcomes was larger for those substances.
That is worth knowing.
Now comes the equally important part.
This study does not prove causation
The design was cross-sectional. Researchers measured reported use and reported outcomes; they did not randomly assign substances or control everything else happening in participants’ lives.
The authors explicitly note that the design cannot establish causality.
Preexisting psychiatric history could differ between groups. People may have used other substances. Respondents were asked to attribute symptoms to a drug themselves, and that attribution could vary depending on how dramatic or physically demanding an experience felt.
The ibogaine and 5-MeO-DMT groups were also relatively small, producing wider uncertainty around estimates.
In short: the outcomes are signals, not verdicts.
Why real-world safety data still matter
Clinical trials deliberately select participants, control dose and purity, screen for risks and provide supervision. Those conditions are features, not flaws, when the goal is to test an intervention.
But they are not the conditions under which every person uses a psychedelic.
General-population data can reveal patterns that highly controlled studies may miss: unsupervised use, uncertain products, combinations with other drugs, medical vulnerabilities and different motives for use.
The price is weaker causal certainty.
This is why safety surveillance needs multiple kinds of evidence rather than one definitive study that settles everything forever.
“Psychedelics” is too broad a safety category
One useful implication of the study is how quickly the umbrella term breaks down.
Ibogaine has a very different physiological risk profile from psilocybin. 5-MeO-DMT produces a very different acute experience from LSD. MDMA is pharmacologically distinct from classic psychedelics even though it often appears in the same therapeutic conversation.
Treating all of them as interchangeable can make both enthusiasm and fear scientifically sloppy.
The better question is always more specific: which substance, which dose, which person, which context, which outcome, over what period?
That is less exciting than “Are psychedelics safe?” It is also a question evidence can actually approach.
The field needs boring honesty
Psychedelic medicine has attracted investment, activism, personal testimony and legitimate scientific interest. That mix creates powerful incentives to simplify.
Positive trial? Breakthrough.
Adverse-event survey? Disaster.
Neither reflex helps people understand risk.
The authors of the new study land somewhere more useful. They argue that the findings support continued safety surveillance, targeted harm reduction where warranted and balanced public-health communication about both potential risks and benefits.
That is not fence-sitting. It is what uncertainty looks like when it has grown up.
If psychedelic science wants public trust, it cannot treat inconvenient data as hostile and it cannot treat every alarming association as proof.
The interesting work is in the middle, where the evidence is usually messier and more informative.