Bottom line up front. I do not certify cannabis for psychiatric indications. Not because patients are lying about the relief they feel, but because the evidence does not support durable functional benefit, THC carries real psychiatric risk, and regular use can manufacture the very anxiety, sleep, irritability, mood, and concentration symptoms patients believe it is treating. This is a long post because the reasoning is layered. If you read only two parts, read "Why patients are so sure it helps" and the closing section "Where I land, and what I do instead." Those two carry the clinical core.
Someone asked in the forum last week whether prescribers have the authority to prescribe marijuana, and what I think about it clinically. It is a deceptively large question, and the honest answer needs both a legal piece and a clinical piece that usually get tangled together. This is my attempt to separate them and then say where I land and why.
If you have followed the news this year, you probably absorbed the headline that marijuana was reclassified from Schedule I to Schedule III. That is the version that reached the public, and it is misleading in a way that matters for how we talk to patients.
What actually happened in April 2026 is narrower than the headline. The federal government moved two specific categories from Schedule I to Schedule III: marijuana contained in an FDA-approved drug product, and marijuana subject to a state medical marijuana license. Everything else, including recreational marijuana and any cannabis outside those two channels, is still sitting in Schedule I right next to heroin. The broader question of whether marijuana as a whole should move is not settled. It is the subject of a DEA administrative hearing, and there is already litigation trying to reverse what was done.
The reason the split looks strange is that it is not a pharmacology decision. It is a legal one. The action was taken under the Attorney General's authority to align scheduling with our obligations under an international drug treaty, which allowed it to take effect immediately without the full rulemaking process, but only for categories that could be wrapped in the controls the treaty requires. Generic "all marijuana" could not be moved that way, so it was left behind. The plant is identical. The legal packaging around it is what differs.
I am starting here because the gap between "marijuana is now Schedule III" and what actually changed is the same kind of gap we are about to see on the clinical side. The simple version that circulates is not the version that survives a close look.
Before any question about efficacy, there is a problem of what we are even talking about. Cannabis is not a drug in the sense we usually mean. The plant contains more than 500 identified chemical compounds, including over 100 phytocannabinoids and roughly 120 terpenes. We have meaningful research on two of them, THC and CBD. The rest are comparatively uncharacterized.
This matters because the entire conversation collapses two very different things into one word. When we say "prescribe marijuana," we could mean an isolated, single-molecule compound, or we could mean a botanical mixture of hundreds of compounds in proportions that vary by strain, grow, and product. Those are not the same clinical object, and reasoning about one tells you very little about the other.
The cannabis industry has a name for why you supposedly cannot just isolate the useful part: the entourage effect, the claim that the compounds work synergistically and the whole plant outperforms any isolated piece. Here is the thing worth sitting with. The mechanism sounds reasonable, but the actual human evidence for it is thin, resting mostly on anecdote and patient report rather than controlled trials.
And here is why that argument backfires on the people making it. The entourage effect is the industry's reason for why you need the whole plant instead of an isolated compound. But if that is true, then you also cannot take the THC out, which means you cannot separate whatever benefit there might be from the psychosis risk and the dependence that THC brings. The benefit and the harm come in the same package, by their own logic. And if the entourage effect turns out not to be true, then there was never a reason to use the whole plant over a clean, isolated compound to begin with. So the claim that is supposed to justify whole-plant cannabis actually argues against using it for a psychiatric patient. Either the harm comes bundled with the benefit, or you never needed the whole plant in the first place, just a single isolated compound from it.
Now the authority question, which has a cleaner answer than most people expect.
You do not prescribe botanical marijuana the way you prescribe sertraline or aripiprazole. There is no dose, no sig, no pharmacy fill. What happens under state medical programs is certification or recommendation. You attest that a patient has a qualifying condition, and they take that attestation to a state-licensed dispensary. The mechanics vary considerably from state to state, and that is worth verifying for your own state before you rely on any of it.
There is a separate, cleaner category you genuinely can prescribe: the FDA-approved cannabinoids. Dronabinol and nabilone are synthetic THC products approved for chemotherapy-induced nausea and AIDS-related anorexia. Cannabidiol as Epidiolex is approved for specific seizure syndromes. These are real prescriptions with defined indications. Notice that not one of them is approved for a primary psychiatric condition. There is no FDA-approved cannabis product indicated for anxiety, depression, PTSD, ADHD, insomnia, or any psychiatric disorder. That absence is itself information about where the evidence actually sits.
Let me go condition by condition, because the strength of the evidence is not uniform and the differences are clinically important. The throughline to watch is the split between isolated CBD, which shows a weak and inconsistent signal with a clean safety profile, and THC or whole-plant cannabis, which carries most of the risk and most of the negative or absent findings. THC is also the compound responsible for the high, though whether it produces that effect on its own or in combination with some of the other 500-plus compounds in the plant is one more thing we do not fully understand.
Anxiety is the most common reason patients ask, and it is the messiest. For isolated CBD, a systematic review of randomized trials found contradictory results across studies but suggested, on balance, that CBD may modestly reduce anxiety with few adverse effects compared to placebo. The mechanism is plausible, with CBD acting as a partial agonist at the 5-HT1A receptor. There are two problems with leaning on that signal, though. First, the doses in those trials ranged from under 15 mg to 800 mg with no consistent protocol, so even the positive finding is not something you could translate into a real prescription. Second, and this is the part worth knowing, those trials were not all using the same thing. Some used purified or synthetic CBD, which is a true single compound, while others used natural CBD-rich preparations that still carry trace amounts of THC and other cannabinoids. The review folded them together. So even when researchers set out to study "CBD for anxiety," the product in the bottle was not always clean CBD, which means the already-weak signal is hard to attribute to CBD alone rather than to whatever else came along with it. And pulling the other direction entirely, meta-analysis shows that cannabis use overall is associated with more anxiety, not less, and the association is stronger for cannabis use disorder. So the purified component may help a little, the trials testing it were not always actually testing it in isolation, and the actual substance most people use is associated with the very thing they are trying to treat.
Sleep is weaker still, and arguably negative for whole-plant cannabis. The strongest recent objective data comes from a controlled trial in diagnosed insomnia patients, where a single oral dose of THC and CBD decreased total sleep time by roughly 24 minutes compared to placebo, with no improvement in time spent awake after falling asleep. Subjective reviews are mixed and limited, and a meta-analysis of randomized trials concluded the evidence for cannabinoids improving sleep quality remains weak. CBN, the cannabinoid currently marketed as a sleep aid, has minimal evidence and in a large trial performed no better than melatonin. So patients feel like it helps them sleep, and the objective data does not support it, and at least one good trial shows it making sleep shorter.
PTSD is the cleanest no despite being one of the loudest yes claims in the culture. State evidence-based guidance, reviewed in 2025, concluded there is insufficient evidence to say cannabis is either effective or ineffective for PTSD. A 2024 systematic review suggested cannabinoids might help specific symptoms like sleep disturbance, but not overall PTSD symptom burden. Be careful with that sleep finding, though, because it is exactly the kind of claim the sleep section above should make you skeptical of. The PTSD signal is based on patients self-reporting better sleep, which is the same subjective measure that does not hold up when you put people on objective monitoring and watch their sleep actually get shorter. So the most favorable read of the PTSD literature is that cannabis does not improve the disorder, and the one place patients say it helps is the place where self-report is least trustworthy. That is not much of a case.
ADHD is where I want to be most direct, because it is the indication patients most often defend on the grounds that it obviously helps them focus. The Canadian ADHD Resource Alliance issued a position statement at the end of 2024 concluding there is no evidence that cannabis is an effective treatment for ADHD or that it improves attention. And it is worth being precise here, because this is not just an absence of evidence that it helps. The evidence points the other way. Cannabis impairs the exact cognitive functions ADHD already compromises, including working memory, processing speed, and sustained attention. So we are not in neutral territory where the jury is out. We are looking at a substance that measurably worsens the thing patients are asking it to fix.
The reason patients are so convinced anyway is a gap between how it feels and what it does, and here is how I understand the mechanism, because it explains the anxiety confusion too. Cannabis does not calm anxiety in the way an anxiolytic does. What THC actually does is disrupt the continuity of attention. You stop being able to hold a single thread of thought for very long, so you move from one thought to the next to the next instead of staying locked on one. For someone whose distress runs through rumination, a worry thought looping on itself, that disruption feels like enormous relief, because the loop breaks. But the loop did not break because the worry resolved. It broke because the person got moved off the thought before they could finish it. My read is that this is the same impairment seen from two sides. To the ruminating patient it feels like calm. On a cognitive test it shows up as inattention and degraded working memory. The relief patients describe and the impairment researchers measure may not be two different effects at all. They may be one effect described by two different observers.
This is also why the restlessness piece is misleading. Part of what patients read as being calmer is the blunting of drive and motivation that comes with regular use, which the literature describes as an amotivational effect. Lower activation can feel like lower anxiety from the inside, but it is not the anxiety being treated. It is the engine being turned down. There is also a second trap specific to ADHD, and it compounds the first. Impulsivity and difficulty delaying gratification are core features of the disorder. When you take a brain that is already wired to act on immediate reward and you introduce a substance that reliably delivers fast relief, you have a setup primed for dependence. The brain learns quickly that cannabis equals relief, and the same impulsivity that defines the ADHD makes it harder to pause, question the pattern, or delay the next use. So ADHD patients are more likely than average to form the habit in the first place. And because they are also the population most prone to misread the in-the-moment calm as genuine improvement, they end up both more likely to use and more confident that it is working. The vulnerability to dependence and the vulnerability to misattribution stack on top of each other.
Psychosis is the harm signal that anchors the whole picture. THC is a recognized risk factor for psychosis, with a dose-dependent relationship, earlier age of onset, and increased risk of transition in people at high risk, modulated by genetic vulnerability. The risk concentrates in adolescents and young adults, in anyone with a personal or family history of psychosis, and in heavy or high-potency use. This is the population where certifying cannabis is not just unsupported but potentially harmful.
This is the part I find most important, and it is the part that almost never makes it into these conversations. If the controlled evidence is this thin, why are so many patients absolutely certain that cannabis treats their anxiety, their sleep, their irritability, their focus? They are not lying, and most of them are not imagining the relief. So what is happening?
The answer is a mechanism called negative reinforcement, and it runs through cannabis withdrawal. Regular cannabis use produces a withdrawal syndrome when it wears off or the person cuts back. And the felt effect wears off relatively quickly. The acute high from smoked or vaped cannabis peaks within about half an hour and is largely gone in two to four hours, which means a daily user is not sitting in a steady state. They are cycling up and back down several times a day, and each comedown is a small withdrawal that the next use relieves. This is the same principle that makes short-acting benzodiazepines like alprazolam so reinforcing and so habit-forming. The shorter the felt duration, the more often the person dips into discomfort and the more often using fixes it, which carves the loop deeper.
One important caveat so the pharmacology is precise. The high being short-acting does not mean THC leaves the body quickly. THC is highly fat-soluble and accumulates in tissue, with a half-life that stretches to many days in frequent users, which is part of why the full withdrawal syndrome can take weeks to settle once someone stops entirely. The subjective effect and the tissue clearance run on two different clocks. The fast one drives the daily reinforcement loop. The slow one drives the long tail of withdrawal.
The DSM-5 criteria for cannabis withdrawal are irritability, anger or aggression, nervousness or anxiety, sleep difficulty including insomnia and disturbing dreams, decreased appetite, restlessness, depressed mood, and physical symptoms like headache or stomach upset. Read that list again with a clinician's eye. It is, almost item for item, a synthetic psychiatric presentation. The withdrawal state manufactures anxiety, insomnia, irritability, low mood, and restlessness, which is the same symptom cluster that defines the conditions people use cannabis to treat.
This is worth pausing on, because it is not unique to cannabis. The way I explain it to patients is that withdrawal from most substances produces a similar nonspecific cluster: disturbed sleep, mood swings, irritability, anxiety. The body has adapted to the substance, and when the substance leaves, the adaptation is briefly unopposed, and that shows up as distress. So the trap of mistaking withdrawal relief for treatment is a general one, and once you see it clearly in one place you start seeing it everywhere.
The clearest parallel is something we see constantly in our own prescribing. A patient on an antidepressant misses several doses, feels noticeably worse within a few days, restarts, feels better within a day or two, and concludes the medication was obviously working for them. But look at the timeline. True therapeutic effect and true relapse do not move that fast. A few days off is far too short for a depressive disorder to genuinely return, and a day back on is far too short for an antidepressant to genuinely treat it, because both directions run on a neuroadaptive timescale of weeks, not days. What the patient actually experienced was discontinuation symptoms coming on fast when the dose was missed, and then resolving fast when it was reinstated. The quick worsening and quick relief feel like proof of efficacy, when the speed itself is the tell that this was withdrawal, not treatment.
The honest way to find out whether a medication is truly helping is to watch what happens past the withdrawal window. Discontinuation symptoms come on quickly but improve with time. So if a medication was genuinely treating the underlying disorder, then once the withdrawal period passes one of two things happens: the disorder stays worse off the medication and does not substantially recover, or, if there was little withdrawal to begin with, the disorder slowly worsens over the following weeks on the same timescale it took to improve in the first place. That slow reemergence, not the fast rebound, is the signal that the medication was doing real work. The same test applies to cannabis, and it is the test almost no one applies, because the fast rebound is so convincing in the moment that the patient never gets far enough past it to see the truth.
Now watch the loop. The patient feels anxiety or insomnia or irritability rising. That rising distress is, for a regular user, often withdrawal. They use cannabis. The symptom resolves. They conclude that cannabis treats their anxiety, when what actually happened is that the cannabis relieved a withdrawal symptom the cannabis itself created. The relief is real. The attribution is wrong. And because the user cannot see the withdrawal process from the inside, the belief that cannabis is medicine gets reinforced every single cycle.
This is not specific to anxiety. It generalizes across exactly the symptoms your patients report, because the withdrawal syndrome generates all of them. Sleep is the clearest example and the most clinically stubborn. Withdrawal-driven insomnia is the most troublesome symptom in early abstinence and a leading reason people relapse, and it can persist for 30 to 45 days after stopping. That long tail is why "I can't sleep without it" feels to the patient like proof of a sleep disorder that cannabis treats, when it is actually evidence of dependence that cannabis produced. The same logic runs through irritability, emotional lability, and concentration. The withdrawal symptom and the target symptom are the same symptom.
The neurobiology backs this up rather than contradicting it. In people with mood and anxiety disorders who use cannabis to self-medicate, imaging shows blunted reward response in the striatum, which is the opposite of what you would expect if cannabis were genuinely treating the underlying condition. And the broader literature is direct about it: the perceived effectiveness of medical cannabis is substantially attributable to coping, to relief of its own withdrawal effects, and to placebo, with no studies finding durable improvement of the underlying disorder.
There is a diagnostic landmine in here that is worth naming for our work specifically. Because the withdrawal criteria are themselves anxiety and depressive symptoms, a daily user at intake presents a picture where you cannot cleanly separate the primary psychiatric disorder from the withdrawal state without a period of abstinence. The substance you are being asked to certify as treatment is simultaneously generating the symptoms you are being asked to treat. That is not a situation where certification makes clinical sense.
So here is my position, offered as reasoning rather than as a rule for anyone else to adopt.
I do not certify or recommend cannabis for psychiatric indications. Not on moral grounds, and not because I think patients are foolish for using it. I do not certify it because I cannot point to controlled evidence that the benefit outweighs the risk for the conditions I treat, because the one component with any real safety and efficacy signal is CBD and that is not what a dispensary is selling, because the thing patients actually use is high-THC product where the evidence is weakest and the psychosis risk is real, and because the substance reliably manufactures the very symptom cluster I would be certifying it to treat. When the strongest argument for the whole plant is an entourage effect that, if true, means I cannot separate the benefit from the harm, the case does not come together.
What I do instead is treat cannabis use as clinical information rather than as a treatment to bless or forbid. If a patient is using, I want to know, in detail and without judgment, because I cannot assess their actual psychiatric baseline without accounting for use and withdrawal. I screen for cannabis use disorder directly. I am especially careful with anyone young, anyone with a psychosis history or family history, and anyone whose "treatment-resistant" anxiety or insomnia might actually be a withdrawal cycle I have not yet untangled. And I have the honest conversation, which is usually more productive than either cheerleading or prohibition: here is what the evidence shows, here is the loop you may be caught in, and here is what it would take to find out what your symptoms actually look like without it.
The standard I hold it to is the same one I hold everything I prescribe to. The goal is never that a patient feels different in the moment. The goal is functional improvement, that they are sleeping, working, connecting, and living better than they were. Feeling something is not the bar. Functioning better is the bar, and it is the bar for medications, for therapy, and for cannabis equally. When I apply that standard, the in-the-moment relief patients describe with cannabis mostly does not translate into the functional gains I would expect from something that was actually treating the disorder.
There is also a quieter piece of evidence sitting right in front of me in almost every one of these conversations. The patients telling me how much cannabis helps their anxiety, their sleep, their PTSD, are, by and large, still in my office asking me to treat their anxiety, their sleep, and their PTSD. If it were genuinely doing the work they believe it is doing, they would not be sitting across from me looking for something more. That is not a gotcha I throw at patients, but it is an honest question worth raising gently, because it often opens the door to the real conversation. If this were treating the problem, why are we still here.
I know colleagues who land in a different place, particularly on a harm-reduction model for patients who are going to use regardless, and that is a defensible position I am not trying to argue anyone out of. But I think it has to be built on the actual evidence rather than on the patient's certainty that it works, because that certainty is, more often than we acknowledge, a symptom of the thing itself.
If you made it this far, the question that started all this was a good one, and worth the long answer. I am glad it came up.
Sources and references are in the first comment below.