ERP Therapy for OCD: The Complete Guide

In 1966, a psychologist at a London hospital tried something that sounded almost cruel. He asked two patients with severe compulsions to face the exact situations they dreaded, and then made sure they couldn’t perform their rituals afterwards. Both improved Meyer, Behaviour Research and Therapy, 1966.

Sixty years later, that experiment has become ERP therapy, exposure and response prevention, the treatment that every major OCD guideline lists first (APA practice guideline, American Journal of Psychiatry, 2007; NICE CG31). It’s also the treatment most people have heard of and least understand. This guide covers what ERP actually is, what the trials found (including the awkward parts), the myth about anxiety needing to “come down”, the single factor that best predicts who gets well, and how to start.

What ERP therapy is, in one table

ERP is a form of cognitive behavioural therapy (CBT) built for OCD. It has two moving parts, and both matter.

Part	What it means	Example
Exposure	Deliberately facing a thought, image, object or situation that sets off the obsession	Touching a door handle; writing “I might have left the stove on”; letting a harm thought sit there
Response prevention	Not doing the compulsion that usually follows, or delaying it and doing it less	No washing; no going back to check; no mental review or asking “are you sure?”

The goal is not to prove that nothing bad can ever happen. It’s to change your relationship with uncertainty: to learn, through experience rather than argument, that you can feel the doubt, skip the ritual, and still be OK.

The key insight is that the compulsion is the problem, not the thought. Everyone has intrusive thoughts. In OCD, the ritual that follows tells the brain the thought was a genuine alarm, so it fires again, louder. Exposure without response prevention (facing the fear, then ritualising anyway) tends not to help. It’s the second half that does the work.

Why compulsions keep OCD going

Here is the loop ERP is designed to break:

1. A trigger appears (a doorknob, a thought, a doubt about your relationship).
2. Anxiety or a “not right” feeling rises.
3. A compulsion (washing, checking, reviewing, asking) brings quick relief.
4. The brain files the sequence as a success: the danger was real and the ritual saved me.
5. Next time, the trigger feels stronger and the ritual feels more necessary.

The relief in step 3 is what makes OCD self-sustaining. The classic explanation, from Foa and Kozak’s emotional-processing theory, is that fear is stored as a structure of “this stimulus means danger” links, and it only updates when you encounter the feared thing and get information that contradicts the danger. Compulsions and avoidance block exactly that information Foa & Kozak, Psychological Bulletin, 1986.

If you want the mechanics of the belief side of this loop, we’ve covered thought-action fusion, the conviction that thinking something is as bad as doing it, and the way reassurance seeking quietly works as a compulsion other people perform for you.

Practice · exercises from the ocd.app app

Try it: plan your first exposure

Face the thought 2 min Spot the OCD thought 2 min Ride the wave 1 min

Exposure and response prevention (ERP) means facing the thoughts OCD makes you avoid, without doing the ritual. Pick a thought from the app that makes you uneasy, then stay with it.

• I must always be certain
• I need absolute certainty
• I must fear uncertainty
• I have to know everything
• Uncertainty is unbearable
• Uncertainty is overpowering
• I can't tolerate not knowing
• Without control - I'm useless
• I need to know everything
• I must have absolute control

Educational practice, not therapy or a diagnosis. More exercises from the app · Practise daily in the ocd.app app →

What the trials actually found (including the awkward bits)

ERP is one of the most studied psychological treatments in existence. Here is what the numbers say, without the marketing.

Against doing nothing, the effect is very large. A meta-analysis of 37 randomised trials found effect sizes of 1.31 for CBT (mostly ERP) versus waiting list and 1.33 versus placebo conditions, which in psychotherapy research counts as very large. CBT also outperformed antidepressant medication (effect size 0.55), and adding medication to CBT did not make it significantly better Öst et al., Clinical Psychology Review, 2015.

Head to head with medication, ERP held its own. In the landmark trial that compared intensive ERP, clomipramine, both together, and placebo, 86% of people who completed ERP were rated as responders, against 48% for clomipramine and 10% for placebo. Adding the drug to ERP did not improve on ERP alone Foa et al., American Journal of Psychiatry, 2005.

When medication isn’t enough, ERP is the recommended next step. In a trial of 100 people whose OCD had only partly responded to an SSRI-type drug, adding ERP produced a response in 80%, compared with 23% for adding the antipsychotic risperidone and 15% for placebo. The authors’ conclusion: offer ERP before antipsychotics Simpson et al., JAMA Psychiatry, 2013. Six months on, the ERP group were still doing significantly better Foa et al., Journal of Clinical Psychiatry, 2013.

Now the awkward part. A 2021 meta-analysis of 36 trials (2,020 patients) confirmed a large effect versus psychological placebo (g = 1.13), but found ERP was not more effective than other active, structured psychotherapies, and only marginally better than properly dosed medication. It also flagged that most trials were run by researchers with a stake in the result, and that the few trials without that “allegiance” showed much smaller effects Reid et al., Comprehensive Psychiatry, 2021.

What to take from this: ERP works, and it works at least as well as anything else we have for OCD. It is not magic, the quality of the evidence is uneven, and how well it works depends heavily on how it’s done, which is what the rest of this guide is about.

ERP is not the only research-backed option, either. In a 2025 trial of 176 adults, group acceptance and commitment therapy held up against group CBT/ERP. Our guide to ACT therapy for OCD explains how it works and how the two fit together.

A second alternative skips planned exposure altogether. Inference-based CBT targets the “maybe” that starts each obsession, and in a 197-person trial people found it easier to stick with than CBT with ERP.

The myth: your anxiety does not have to “come down”

For decades, ERP was taught as a habituation exercise. You stayed in the exposure until your anxiety dropped, ideally by half, and that drop was the proof it had worked. Many therapists still teach it this way, and many people quit because their anxiety stubbornly didn’t drop.

The current model, called inhibitory learning, says something different. Exposure works by building a new memory (“I touched it and nothing happened”) that competes with the old fear memory. What matters is how surprised you are, not how calm you feel. Researchers call this expectancy violation: you predict “if I don’t wash, I’ll be sick with anxiety all day”, you don’t wash, and the day turns out to be survivable. That mismatch is the learning. Anxiety dropping during the session is nice but optional Craske et al., Behaviour Research and Therapy, 2014.

This changes how good ERP is run:

• Before an exposure, you name a prediction. “If I leave without checking, I’ll be 80% sure the house will burn down.” Afterwards, you compare.
• You vary the exposures (different places, times, moods) rather than repeating one until it feels easy, because varied learning generalises better.
• Safety signals get removed. Doing an exposure with your phone in hand “just in case” teaches you the phone kept you safe.
• A tough session is not a failed session. If you did the exposure and skipped the ritual, you did the work, whatever your anxiety did.

If you’ve tried ERP before and it “didn’t work”, ask yourself whether it was run as habituation and you got stuck waiting for calm that never came.

What happens in ERP therapy, session by session

A typical course is around 12 to 20 sessions. Research protocols often use 17: a few introductory sessions and 15 exposure sessions Simpson et al., Journal of Consulting and Clinical Psychology, 2011. Here’s the shape of it.

1. Assessment and mapping (sessions 1 to 3). Your therapist maps your obsessions, your compulsions (visible ones and mental ones), your avoidance, and the situations you have quietly rearranged your life around. You’ll usually complete a severity scale like the Y-BOCS so progress can be measured.

2. Building the hierarchy. Together you list feared situations and rate each from 0 to 100. This becomes your practice plan. We explain how one is built, and why the hardest items don’t come first, in our guide to the ERP hierarchy.

3. In-session exposures. You start somewhere in the middle of the list, not the bottom. The therapist does the exposure with you, coaches you through the urge to ritualise, and helps you notice what actually happened versus what you predicted.

4. Response prevention rules. You agree specific rules for the week: “one lock check, then leave”, “no re-reading sent emails”, “no asking my partner if I’m a good person”. For mental compulsions, the rule might be “notice the review starting and return to the task”, which we cover in Mental compulsions in OCD.

5. Homework. Between sessions you repeat the exposures on your own, daily, and log them. This is where most of the change happens, and most of the dropouts (more on that below).

6. Moving up the list, then relapse prevention. As items lose their charge you move up. The final sessions cover what to do when OCD tries a new theme, which it usually does.

For what to ask a prospective therapist and how to tell an OCD specialist from a general one, see OCD therapy: what to expect.

The one factor that best predicts who gets well

Here’s the finding that should change how you approach ERP. Two studies from the same research group tracked how well people stuck to their between-session assignments, and what happened to them.

In the first, therapists rated homework adherence at every session for 30 adults. Higher adherence predicted lower OCD severity at the end, and early adherence (sessions 5 to 9) was already predictive. Every other predictor the team measured, including baseline severity and the quality of the therapy relationship, worked through adherence Simpson et al., Journal of Consulting and Clinical Psychology, 2011.

In the second, with 37 patients, the team split adherence into three pieces: how many homework exposures you attempted, how well you did them, and how successfully you resisted the ritual afterwards. All three mattered, but success with response prevention was the strongest predictor, and it identified early in treatment who would finish well Wheaton et al., Behaviour Research and Therapy, 2016.

In plain terms: the sessions teach you the method, but the daily practice between them is the treatment. That is also why structured between-session practice is the entire point of the ocd.app program, and why a therapist who never asks about your homework is a warning sign.

Weekly, intensive, four-day, online: which format works?

ERP comes in several shapes, and the evidence for each is different.

Standard weekly or twice-weekly sessions are what most trials tested and most clinics offer.

Intensive ERP compresses treatment into daily sessions. The Foa trial above used four weeks of intensive sessions. In a 2026 naturalistic study we covered in our September research roundup, a high-intensity programme (ERP three times a week plus adherence tracking and family involvement) reached remission in 67.6% of patients Levy et al., Comprehensive Psychiatry, 2026.

The Bergen 4-Day Treatment takes concentration to its limit: ERP delivered over four consecutive days in a small group. In the first randomised trial, 93.8% of the four-day group responded and 62.5% were in remission, versus 12.5% and 6.3% for a self-help condition and 0% on a waiting list. Nobody dropped out Launes et al., Frontiers in Psychology, 2019.

Video teletherapy. The largest treated cohort ever reported, 3,552 adults doing ERP over video with app-based tools between sessions, saw a 43.4% average symptom reduction and a 62.9% response rate, using a mean of just 10.6 therapist hours. Gains held at 12 months Feusner et al., Journal of Medical Internet Research, 2022.

Internet-delivered CBT. A Swedish randomised trial compared therapist-guided internet CBT, unguided internet CBT, and face-to-face CBT. Guided online treatment came close to face-to-face (the difference was 2.1 Y-BOCS points, statistically inconclusive) and was cheaper. Unguided treatment was clearly less effective Lundström et al., JAMA Network Open, 2022. The lesson: online works, but a human in the loop matters.

Virtual reality. Adding six VR exposure sessions to usual care did not beat usual care alone in a 2026 trial of 80 people with contamination or checking OCD, though patients liked it Rolvien et al., Journal of Medical Internet Research, 2026. Promising, not proven.

“Isn’t ERP too brutal? Doesn’t everyone quit?”

This is the fear that stops most people from starting, so it deserves data rather than reassurance. A review of 21 randomised trials (1,400 participants) found a weighted dropout rate of 14.7% for ERP, statistically no different from other OCD treatments such as cognitive therapy Ong et al., Journal of Anxiety Disorders, 2016. Roughly one in seven people leave early, which is in line with psychotherapy generally.

Exercise after exposure. In a 2026 randomised trial with 97 inpatients, people who did high-intensity exercise straight after each ERP session improved more than a low-intensity comparison group, and the gap held at six months (details in our October 2026 OCD research roundup). One trial, but an intriguing add-on to ask about.

Two things keep it tolerable. First, ERP is collaborative: you choose the items, the pace, and the order. Nobody should ever spring an exposure on you. Second, good ERP works with reasonable risk. Washing your hands after the toilet is hygiene; washing for forty minutes until it feels “right” is a compulsion. ERP targets the second, never the first.

ERP does need extra care, and usually a specialist, if you’re also dealing with severe depression or suicidal thoughts, trauma symptoms, active substance use, or a medical condition that limits which exposures are safe. Those aren’t reasons not to treat OCD. They’re reasons to have the plan made by someone who knows what they’re doing.

Does it last?

Better than most people fear. Across 16 trials, CBT for OCD still outperformed control conditions at follow-up, though with a smaller effect than immediately after treatment Olatunji et al., Journal of Psychiatric Research, 2013. In the SSRI-augmentation trial above, gains held across six months of maintenance, and the people who improved most during treatment were the ones who kept their gains Foa et al., Journal of Clinical Psychiatry, 2013.

The practical implication: OCD tends to return under stress or with a new theme, and the skill you keep is knowing what to do when it does. Our OCD relapse prevention guide covers that stage, and if family members have been pulled into your rituals, family accommodation explains how they can step back without a fight.

How to start

• Get a baseline. If you’re not sure whether this is OCD, the free OCI-4 screening test takes two minutes and gives you a score to bring to a clinician.
• Look for the words “ERP” or “exposure and response prevention” in a therapist’s profile. “CBT for anxiety” is not the same thing. The IOCDF directory lists specialists.
• Ask how they run exposures. “How do you pace it?” and “How do you handle homework?” tell you a lot.
• Practise daily. Whatever else you do, this is the variable the research keeps pointing at. Our OCD exercises page is a starting point for structured daily practice.

Frequently asked questions

Is ERP the same as CBT?

ERP is a specific type of CBT developed for OCD. General CBT focuses on identifying and challenging thoughts; ERP focuses on facing triggers and dropping compulsions. Research comparing ERP with cognitive therapy for OCD finds similar results, and most modern programmes blend the two.

How long does ERP therapy take?

Most protocols run 12 to 20 sessions over three to five months, with daily practice between sessions. Intensive formats compress this into weeks, and the Bergen 4-Day Treatment into four days. People often notice change within the first month, but timelines vary with severity and how consistently the homework gets done.

Does ERP work for Pure O or mental compulsions?

Yes. Mental rituals (reviewing, neutralising, self-reassurance) are compulsions, so response prevention applies to them too. Exposures are often imaginal: writing out the feared thought or scenario and sitting with it without analysing.

Can I do ERP on my own?

Self-directed practice can help with milder symptoms, and structured tools improve consistency. Trials suggest that having a clinician in the loop, even remotely, produces better results than fully unguided programmes, so use self-help to support treatment rather than replace it, and get professional guidance for severe or complex OCD.

What if ERP feels too scary?

ERP is paced and collaborative. You choose the starting point and nobody should force an exposure. If a therapist is pushing you into your hardest fear in week one, that is poor ERP, not a sign that you can’t do it. A tough session where you skipped the ritual still counts as progress, however your anxiety behaved.

If you want to try that first step now, we’ve put together seven beginner ERP exercises you can do at home, with the prediction-and-compare routine that makes each one count.

---

This article is for educational purposes and is not a substitute for professional diagnosis or treatment. If you’re in crisis, contact your local emergency services or a crisis line.