Good AI CASC feedback should identify what the learner actually did, connect it to the station's task and suggest a specific improvement that can be tested in another attempt. Fluency, politeness and a confident summary are not enough. Human observation remains important where the relevant behaviour, interaction or physical performance is not adequately captured by the simulator.
This article is published by iatroX and includes its simulation design as a comparator where the capabilities are relevant. It is a public-documentation review dated 24 September 2026, not a hands-on assessment of MRCPsych Mentor or evidence that either platform predicts CASC performance.
What MRCPsych Mentor currently advertises
MRCPsych Mentor's website, checked on 24 September 2026, advertises an AI-powered CASC simulator and describes alignment with the RCPsych curriculum and mark scheme. That establishes the provider's stated product purpose. It does not establish College endorsement, equivalence to the actual assessment or independently demonstrated improvements in pass rates.
The Royal College of Psychiatrists' preparation resources distinguish CASC preparation from the written examinations. A bank that helps with Paper A or Paper B can support underlying knowledge, but it does not by itself show how a learner performs an interactive task.
The buying question is therefore not simply whether a simulator exists. It is whether the interaction and feedback address the particular behaviour you need to improve, and what additional practice remains necessary.
A successful conversation is not automatically a successful station
Consider a fictional station in which a person wants to discuss worries about returning to work after a period of illness. A learner may produce a warm, articulate explanation yet never establish what the person is worried will happen. Another learner may gather relevant information but fail to explain a plan clearly.
Those are different performance problems. Generic praise for empathy or a long list of suggested questions may not distinguish them. Feedback should show how the learner's actual choices helped or obstructed the requested task.
A useful debrief separates task understanding, relevant information gathering, organisation, responsiveness and explanation. It also identifies where the available transcript or recording is insufficient to judge something. A system that confidently assesses an unobserved behaviour can create false reassurance.
The fictional example is a communication exercise, not a clinical management protocol or a reproduced College station.
A simulation-feedback inspection rubric
This is an original inspection framework, not a validated marking instrument. Use descriptive observations rather than inventing a numerical pass score.
| Domain | Evidence to look for | A weak feedback pattern | A more useful feedback pattern |
|---|---|---|---|
| Task interpretation | Whether the learner addressed the station request | Generic advice unrelated to the task | Identifies the point at which the discussion drifted |
| Information gathering | Questions actually asked and responses explored | A universal checklist with no transcript reference | Links an omission to its consequence in this case |
| Organisation | Signposting, sequencing and use of time | Praise based only on a tidy final summary | Shows where a topic was repeated or left unresolved |
| Patient concerns | Response to the person's stated priorities | Credits empathy from a stock phrase alone | Examines whether the concern changed the conversation |
| Explanation | Clarity and adaptation to understanding | Rewards volume of information | Identifies an unexplained term or missed understanding check |
| Plan discussion | Connection between information and proposed next steps | Inserts an ideal plan without analysing the attempt | Explains the reasoning gap in the learner's proposal |
| Uncertainty | Recognition of missing information and limits | Treats every judgement as certain | States what cannot be assessed from the interaction |
| Remediation | A specific behaviour to practise next | Advises the learner simply to be more confident | Prescribes a focused retry with a changed scenario |
A demonstration that produces a detailed report does not automatically meet these standards. Inspect whether each comment is supported by the interaction, not merely whether the report is long.
Does the feedback respond to what you actually said?
After a practice station, select one positive comment and one criticism. Find the relevant moment in the transcript or recording. Can you point to the behaviour the system appears to be evaluating?
Suppose the feedback praises exploration of the person's work concerns. In the fictional interaction, the learner asked whether work was stressful, received a hesitant answer and immediately moved on. The comment may overstate what was actually explored. Conversely, a generic omission list may criticise a topic that was already addressed using different wording.
This exercise is not about catching a product out through adversarial prompts. It is about checking whether the report is an accurate account of the learning encounter. Where it is not, record the discrepancy and use the provider's feedback route rather than silently incorporating it into your study plan.
A transcript is also incomplete evidence for some behaviours. Text alone cannot establish eye contact, physical examination technique or the full effect of pacing and tone. Audio adds information, but does not eliminate every observation gap.
Repeat practice without rehearsing a script
Repeating the same opening, questions and closing can improve familiarity while concealing weak responsiveness. A meaningful variation changes the person's concern, the order in which information appears or the way a question is answered, while retaining the learning objective.
For the fictional work-related discussion, the first version might centre on confidence, the second on a misunderstanding of what colleagues know and the third on uncertainty about the next practical step. The learner should not deliver an identical explanation in each version merely because the station theme is familiar.
After feedback, choose one behaviour to improve. For example, explore a concern before offering reassurance, or explain why a particular question is being asked. Then attempt a new version without coaching and inspect whether that behaviour occurs naturally.
The purpose is transfer, not a rising score obtained by learning the simulator's preferred script. A repeated commercial station score should not be treated as a national examination prediction.
What still needs a human?
A peer can observe whether the conversation feels responsive, whether an explanation is understandable and whether important concerns were ignored. A faculty member can add experienced judgement about the task and the appropriateness of the interaction. Neither should merely repeat a checklist while ignoring what happened.
Where physical examination or another observed skill is relevant, it requires appropriate supervised practical assessment. Voice or text practice cannot establish that the learner performs the physical task competently.
A useful peer session has a defined focus. The observer records the behaviour, its effect and one proposed change. "You sounded nervous" is less actionable than "You interrupted the answer before the concern became clear". The learner then repeats a short segment or tries a different case.
Human feedback can also be inconsistent. The answer is not to assume that a person is always right and a simulator always wrong, but to relate feedback to observable behaviour, the stated task and current authoritative preparation guidance.
Where iatroX's simulation design is relevant
Per iatroX product information, September 2026, iatroX Simulations launched with eighteen examination-specific tracks and 1,114 clinician-reviewed cases, supporting voice and text, coached practice, uninterrupted exam mode and transcript-linked domain feedback. These are whole-product launch figures, not counts of CASC cases or evidence of outcomes in psychiatric trainees.
The current public information reviewed for this article did not establish a CASC-specific iatroX track sufficiently to make a direct replacement claim. Candidates should confirm the exact examination option rather than infer it from general psychiatric questions or the overall simulation library.
The relevant comparison is the design principle: feedback should be linked to the attempt, remediation should address a demonstrated gap and later practice should test independence. Clinician review describes a content process; it is not endorsement by the Royal College or proof that every AI response is correct. The iatroX examination catalogue and its dedicated simulation methodology article should be consulted for the scope of available learning functions.
Decide by the missing practice modality
MRCPsych Mentor's advertised CASC simulator is worth inspecting when you need accessible repeated conversation practice and task-related feedback. A study partner or faculty-led session is especially relevant when the missing information concerns observed interaction, nuanced communication or practical performance. Written questions remain useful when the problem is underlying knowledge rather than station delivery.
Use iatroX only for currently available functions that address the specific gap, not on an assumption of CASC equivalence. Per its September 2026 offer, one complete simulation is free; paid simulations are included with question banks, Tutor, study planning and CPD tools rather than sold as a separate add-on. Examine the relevant track before choosing a subscription.
The strongest preparation plan uses each modality for what it can actually reveal. It does not ask a fluent conversation to prove a competence it has not observed.
Frequently asked questions
Is MRCPsych Mentor's AI CASC simulator endorsed by the College?
Its public page describes curriculum and mark-scheme alignment. That claim should not be read as Royal College endorsement or independent validation of equivalence.
Can an AI station score predict whether I will pass CASC?
Not without appropriate outcome-linked validation of that specific score and product. No such predictive claim is established by this review.
Can voice or text simulation replace peer practice?
It can support repeated independent rehearsal, but it does not replace every form of observed feedback or practical skills assessment. Use human observation for the behaviours the simulator cannot adequately capture.
Explore available examination practice and learning tracks →