OCD Treatment

TMS vs Ketamine
vs Psilocybin
For OCD

Of TMS, ketamine and psilocybin for OCD, only TMS is part of standard care, and it's a third-line option for severe OCD that hasn't responded to ERP and medication. Ketamine and psilocybin are still experimental, for research settings only. That's the position of a 2026 BMJ review and the International OCD Foundation's guidelines. None of the three replaces ERP.

I get why people ask. If you've done the hard work and you're still stuck, a new option that promises fast relief is going to sound good. So here's what the evidence actually shows, without the hype in either direction.

How do TMS, ketamine and psilocybin compare?

Does TMS work for OCD?

TMS has modest evidence for OCD. Repetitive transcranial magnetic stimulation uses magnetic pulses through the scalp to change activity in brain areas linked to OCD. Across meta-analyses of randomized trials against sham, the effects were small to medium, and it's considered safe. The catch is that studies use different targets, frequencies and numbers of sessions, so the best way to do it is still unclear (Abramowitz et al., 2026).

The IOCDF places TMS as a third-line option for severe, treatment-resistant OCD and notes that certain forms are FDA-approved for OCD in the United States (IOCDF). A milder cousin, transcranial direct current stimulation, showed no advantage over sham in a meta-analysis of six trials.

Does ketamine work for OCD?

Ketamine looks interesting for OCD, but it's early. Small studies found symptoms dropped within hours of an IV dose and stayed lower for up to about a week. The proposed mechanism involves glutamate, a different brain system from the serotonin system SSRIs work on. What's missing is larger, longer trials, and the IOCDF lists it as experimental (Abramowitz et al., 2026; IOCDF). A week of relief is real relief. It's not the same as learning, and learning is what makes OCD loosen its grip for good.

Does psilocybin work for OCD?

We don't know yet whether psilocybin works for OCD, but the first randomized trial is in. In 2026, researchers gave 15 people with OCD four weekly sessions of high-dose psilocybin, low-dose psilocybin or an active placebo (lorazepam), then four more high-dose sessions for everyone. Psilocybin reduced symptoms and the placebo didn't. After eight weeks, 73% met the bar for response and 40% for remission, and the gains faded but held up substantially at six months, with no serious side effects (Moreno et al., 2026).

Here's how to read that. Fifteen people is tiny, five per group in the blinded part, and the second phase wasn't blinded at all, so the 73% comes from a stretch where everyone knew they were getting the real thing. It's a reason to run bigger trials, not a reason to book a retreat. The full breakdown is in OCD research updates. Earlier, the only published study was an open-label one with nine participants (Abramowitz et al., 2026).

The big research problem is blinding. A psychedelic is hard to mistake for a placebo, which can inflate how well it seems to work. The IOCDF is direct about it: people with OCD shouldn't use psychedelics as treatment outside controlled research (IOCDF). If you're curious, the right route is a clinical trial, not a retreat.

What about deep brain stimulation?

Deep brain stimulation is the most invasive option: an implanted device sends electrical pulses to a specific brain region. It's reserved for adults with severe OCD that hasn't responded to anything else. The IOCDF reports substantial relief for many patients, while the BMJ review notes most studies are small and uncontrolled, and care is limited to specialty centres (Abramowitz et al., 2026; IOCDF).

Should you try these before ERP?

No. Every guideline puts ERP and medication first, and many people who say ERP didn't work never got real ERP. Before you look at magnets or mushrooms, check the basics on treatment-resistant OCD, and see the whole ladder on OCD treatment in 2026: what works. If you're in Ontario and want to talk it through, start with OCD treatment in Toronto.