ERP for OCD
Practise exposure and response prevention for OCD, from rationale to relapse prevention
One patient, Tobias, and a whole treatment: mapping his rituals, the rationale, the hierarchy, exposures with ritual prevention and the ending. He is AI-simulated: an AI language model plays him from a case written by clinicians, and the case carries forward between sessions.
Browsing is open. Starting a conversation needs a free account, and signing up takes under a minute.
Who you can meet
- Tobias, 36
- Leila, 47
- Paulina, 33
- Samuel, 24
How it works
- Choose a patient and listen Each one has a different form of OCD: cleaning, frightening thoughts after a birth, religious doubt, checking.
- One session, or the whole treatment Start at session 1 and take the patient through the treatment, or step into the session you want to practise. Aim for ten replies or more in each.
- Read the feedback Notes on the areas that session is judged on, with the moments in your own conversation that show them.
A whole treatment, session by session
- Session 1: Assessment and rationale
- Session 2: The hierarchy and the plan
- Session 3: The first exposure
- Session 4: The first imaginal exposure
- Sessions 5–16: Continuing exposure
- Session 17: Review and relapse prevention
A ritual record after session 1, then daily exposure without rituals.
Advanced and Expert bring a worse week, or a ritual quietly swapped for another.
What is exposure and response prevention?
Exposure and response prevention (ERP) is the main psychological treatment for OCD. Step by step, from a graded list, the patient approaches what sets off the obsession and refrains from the rituals that usually bring relief. Repeated without the ritual, the patient learns that the feared outcome does not come, or can be borne, and the urge fades.
CBT therapists and psychologists who treat OCD, or are learning to.
What you practise
- Map obsessions, feared consequences, visible and covert rituals, avoidance and reassurance sources.
- Build the ERP rationale from the patient's own examples and check that it landed.
- Anchor the distress scale and build a specific, graded hierarchy.
- State ritual-prevention rules completely and hold them through negotiation.
- Conduct in vivo exposure: encourage approach, stay in it, track distress, no reassurance or distraction.
- Conduct imaginal exposure with a present-tense script that includes the feared outcome.
Feedback while you practise, and on the whole conversation
During the conversation you get a short tip after your replies on what to keep and what to try next. After the conversation, with an account, an AI supervisor reads the whole conversation and writes you a review: one thing to practise, moments replayed with a reply to try instead, what to keep and rethink, how the patient reacted, and what to read.
What the feedback looks at
- Functional assessment: Mapping the obsessions, feared consequences, visible and hidden rituals, avoidance and sources of reassurance.
- Treatment rationale: Building the loop that keeps OCD going from the patient's own examples, and checking that it landed.
- Hierarchy and distress scale: Anchoring the 0 to 100 scale in the patient's life and turning avoidance into a graded list.
- Ritual-prevention rules: Stating the rules completely, holding them through negotiation, and treating violations as information.
- In vivo exposure: Setting up an exposure and staying in it without reassurance or distraction, paced to the patient's ratings.
- Imaginal exposure: A present-tense script with the worst sentence in it, held without neutralising.
- Reassurance and accommodation: Recognising reassurance seeking in its disguised forms, and declining it warmly.
- Processing after exposure: Asking what was predicted and what happened, and letting the patient's own learning emerge.
- Homework and self-monitoring: Setting the work between sessions concretely: what, when, how often, and how it is written down.
- Relapse prevention: Telling a lapse from a relapse, and making self-directed exposure a rule of life.
- Safety and boundaries: Assigning only situations that are safe but avoided, and never going beyond what was agreed.
Read the feedback as suggestions to think over. An AI model can misread a moment, so weigh it against your own judgement and take what matters to a supervisor or colleague.
Built on: PsychBase's own dimensions, written from the UCL CBT competence framework (Roth and Pilling) and the therapist guide by Foa, Yadin and Lichner (2012).
The feedback is produced by an AI model: training feedback, not a certificate of competence.
Deliberate practice: patients who adapt to your level
The lab follows how your replies land. When it goes well it suggests the next level, and when it gets hard, an easier one. A conversation you leave is kept for 30 days, your level is saved for each patient, and your reviews are saved so you can compare with last week.
Clinical lead: David Boardman, MD, psychiatrist
Frequently asked questions
- What does ERP practice cover here?
- A whole treatment: mapping obsessions and rituals, the rationale, the distress scale and hierarchy, the first exposures with ritual prevention, the middle sessions, and the ending.
- What does the feedback in the ERP lab look at?
- Eleven areas, among them the functional assessment, ritual-prevention rules, how exposures are conducted, reassurance and accommodation, and relapse prevention. The dimensions are PsychBase's own, written from the UCL competence framework and the Foa, Yadin and Lichner therapist guide.
- Does the ERP lab rate the patient's symptoms?
- No. The feedback is about how you conduct the treatment. The lab does not score the patient's symptoms, and no symptom scale is reproduced in it.
- Can I practise a single ERP session without doing the whole treatment?
- Yes. Step into any session, for example the first exposure, and practise that part on its own. It is the same case, and the sessions before it count as done.
- What happens if I reassure the patient?
- Much what happens in real treatment: the doubt is answered for a moment and the urge to check comes back. The feedback marks reassurance, including its disguised forms.
This lab is for clinical training. The patient is generated by an AI language model from a written case; no real patient and no patient data are involved. Browsing is open; starting a session needs a free account.
For clinical decision support only. Not a substitute for professional judgment.