Clinitalk is aimed at GP trainees who are already consulting and want structured feedback on their real practice ahead of the SCA. The most important thing to understand before you buy is that it is not primarily a scripted AI-patient simulator: its core tool records your actual consultations and returns AI feedback mapped to SCA criteria, alongside a small free case library and an optional examiner-led course. That makes it strong on skills MCQ banks miss — but its principal limitation is that it grades the consultations you already do rather than manufacturing unseen exam cases on demand.
What Clinitalk offers for MRCGP SCA right now
Checked against the vendor's public pages on 19 July 2026, with all figures vendor-reported. Clinitalk's central feature records audio or video of your real GP consultations — face-to-face or telephone — and returns "immediate individualised SCA feedback" using AI: traffic-light indicators, guideline-adherence checks, and commentary mapped to SCA assessment criteria (for example, flagging where you elicited a patient's ideas, concerns and expectations but did not then address them). It also publishes a free SCA case library — around twenty cases at last check — with interactive mark sheets aligned to the SCA mark scheme, role-player notes of the kind used in the real exam, and a case-quiz AI, all attributed to former RCGP examiners and tutors. An optional one-day online course is run by RCGP examiners. Pricing ran, at last check, from a low-cost "Starter" tier (around £5/month, weekly recording allowance) to a "Standard" tier (around £17/month, daily allowance) and an annual bundle (around £295, course plus twelve months' app access); a free trial is offered to trainees, access is free to trainers, and deanery reimbursement is available in some approved areas. Verify current tiers, allowances, the case-library count and whether any library cases use voice AI on the product page.
The exam your practice must answer to
The SCA is twelve simulated twelve-minute consultations, 144 minutes in total, sat remotely from ST3, nine diets a year, fee around £1,207. Role-players consult by video or telephone through the RCGP platform; you gather a history, ask for examination findings verbally, and articulate management aloud. Each case is marked by a calibrated examiner across three domains — Data Gathering and Diagnosis, Clinical Management and Medical Complexity, and Relating to Others. The official RCGP candidate guidance, sample cases and marking pages define the format; Clinitalk's mapping of its feedback to "SCA criteria" is a helpful third-party interpretation, not the RCGP's own marking, and should be read that way.
Format map: which SCA tasks Clinitalk reproduces
Because Clinitalk works differently from a scripted simulator, its format map looks different too — and in one respect it reaches something the others cannot.
| SCA task | Reproduced by Clinitalk? | Where, and the caveat |
|---|---|---|
| A patient to consult with | Your real patients | Not a scripted role-player; requires you to be in clinic |
| 12-minute timed consultation | Indirectly | You choose which real consultations to record; timing is yours to impose |
| Spoken history and data gathering | Yes, on real cases | Feedback is on genuine consultations — high ecological validity |
| Verbal management planning | Yes | Guideline-adherence checks flag currency |
| The three examiner domains | Partly | Feedback maps to SCA criteria; interpersonal judgement remains hard to automate |
| Interpersonal rapport and non-verbal signals | Closer than most | Because it analyses a real consultation, not a script — but still inferred by a model |
| Unseen exam-style cases at volume | Limited | The free library is small; the tool is not a case generator |
| A full 12-case circuit | No | It is a feedback tool, not a mock-circuit engine |
The distinctive contribution is in the middle rows. A scripted AI patient can never be more than a script; a recording of you actually consulting a real, undifferentiated patient captures your genuine data gathering, your real safety-netting and your true manner. That is precisely the material MCQ banks cannot touch — and it is why, for a trainee already seeing patients, Clinitalk tests something no question bank and few simulators reach.
Fidelity test: timing, interface, scoring and permitted actions
On timing, Clinitalk does not impose the twelve-minute clock; you must record consultations you have deliberately kept to length, which is a discipline as much as a limitation. On interface, it is not trying to mimic the RCGP exam screen at all — it is a feedback layer over your normal clinical setting, so it trains real behaviour rather than a platform. On scoring, its traffic-light and guideline-adherence output is genuinely useful for the observable half of a consultation, and the guideline checks add a currency signal most simulators lack; but its judgement of rapport and reasoning is still model-inferred and needs calibration. On permitted actions, because these are real consultations, everything you can do in clinic is "permitted" — which means the tool cannot enforce the SCA's specific constraint that management is stated, not executed, so you have to impose the exam's verbal-only discipline yourself.
Case-mix audit: is the spread realistic?
Here the ordinary logic inverts. With a scripted simulator you audit whether the library's spread is realistic; with Clinitalk, your case mix is whatever walked through your surgery door, which is both a strength and a risk. The strength is authenticity — real morbidity in real proportions. The risk is that your clinic may under-represent exactly the areas the SCA samples deliberately: paediatrics, women's and sexual health, mental health, ethically loaded and safeguarding consultations, and rarer presentations may simply not present in a given fortnight. So audit your recorded set against the RCGP's published case content and consciously seek out, or role-play, the groups your real list under-supplies. The blueprint discipline in completion is not coverage applies directly: authenticity does not guarantee breadth.
Feedback audit: observable, inferred and generated
Clinitalk's feedback, like any automated feedback, contains three layers. Observable behaviours — did you set an agenda, screen red flags, safety-net, check understanding — are checkable against the recording and reliable, and the guideline-adherence layer usefully extends this into whether your plan was current. Inferred competence — the model's read of your clinical reasoning — is directional. Model-generated commentary is the least reliable, and the interpersonal read is the hardest of all, because even with a real recording a model is estimating rapport from words and prosody rather than judging it as a human examiner would. Trust the observable and the guideline checks; investigate the inferred; hold the generated rapport commentary lightly and confirm it with a human observer. The mechanics are in our pillar on calibrating automated feedback before you trust the score.
Repetition risk and preserving unseen cases
Clinitalk's repetition risk is unusual: because you record real consultations, you rarely repeat an identical case, so the false-fluency trap of a memorised library is smaller. The free case library, however, is finite and small, so if you lean on it for exam-style rehearsal, treat those twenty-odd cases as you would any limited set — do not over-drill them, and preserve a few unseen for late calibration. The larger risk is subtler: recording only your comfortable consultations. If you self-select the encounters you feel went well, your feedback flatters you. Record the messy, undifferentiated and difficult consultations too, because those are the ones the SCA is built from.
What it cannot test
Clinitalk cannot supply examiner variability — it is your patients and one AI model, not a calibrated panel across a diet. It cannot manufacture the specific unseen cases the RCGP will present, so it is not a substitute for scripted-case rehearsal or a full mock circuit. It depends on you already having patients to record, which makes it a poor fit for a trainee between clinical placements or preparing far in advance. And, like every automated tool, it cannot deliver the human judgement of rapport that the Relating to Others domain ultimately rests on. None of this is a flaw in what Clinitalk does; it is the boundary of a feedback tool as distinct from a simulator, and naming it honestly is the point of this audit.
Where iatroX honestly sits
iatroX is not a consultation simulator and does not replace Clinitalk's feedback on your real consultations — that must be said plainly. iatroX sits underneath: it is the clinical-knowledge and unseen-MCQ layer. When Clinitalk's guideline-adherence check flags that your management drifted from current practice, iatroX is where you close that gap and then test, on unseen questions, whether the corrected knowledge holds. It measures the medicine, not the manner. And because a practice score is not a readiness verdict, keep the caveat from why your Q-bank percentage is not your exam score in view: iatroX tells you whether your clinical knowledge is sound, while Clinitalk tells you how you consulted.
A seven-day pattern for busy ST3 trainees
This plan uses Clinitalk for one defined job — feedback on your real consulting — and iatroX for the separate job of measuring clinical knowledge on unseen items, with no claim to any proprietary algorithm. Monday: record two real consultations kept to twelve minutes; review the traffic-light and guideline feedback for observable misses. Tuesday: deliberately consult, and record, a case type your list under-supplies. Wednesday: a knowledge session in iatroX's MRCGP SCA area on the guideline gaps Clinitalk flagged. Thursday: two Clinitalk case-library cases with their mark sheets, plus the case-quiz AI. Friday: peer-observed role-play of a difficult communication case, human feedback compared with the model's. Saturday: record a longer surgery and review by domain. Sunday: rest, preserving unseen library cases. Clinitalk observes the real consultation; iatroX keeps the medicine inside it current.
Continue, supplement, switch or stop
Continue while you are consulting regularly and the feedback is sharpening your observable behaviours and guideline currency. Supplement with a scripted-case simulator or a mock circuit, because Clinitalk does not generate unseen exam cases at volume or run a full twelve-case mock. Switch your reliance if you are not currently seeing enough patients to record — the tool needs a live clinic to work. Stop treating the small free library as sufficient case rehearsal; it is a taster, not a bank. Decide by measured need — clinic access and case breadth — not by novelty or sunk cost.
Three mistakes this audit is designed to stop
First, expecting a scripted AI-patient simulator and being surprised it grades your real consultations instead — Clinitalk's whole value is that it works on genuine encounters, so use it as the feedback layer it is. Second, recording only your good consultations; self-selected material produces flattering feedback and hides the difficult cases the SCA is built from. Third, trusting the interpersonal commentary as if it were an examiner's verdict — even on a real recording, rapport is inferred by a model, so calibrate it against a human observer before you act on it.
Frequently asked questions
Is Clinitalk enough for MRCGP SCA on its own? For a trainee consulting regularly it can carry the consultation-feedback half of preparation well, because feedback on real encounters reaches skills no MCQ bank can — but it does not generate unseen exam cases at volume, run a full mock circuit, or measure your clinical knowledge on unseen questions, so it needs a scripted-case simulator and a knowledge source alongside it.
Which MRCGP SCA component does Clinitalk not reproduce well? The manufacture of unseen, exam-style cases and a full twelve-case timed circuit: Clinitalk is a feedback tool on your own consultations, not a case generator or a mock engine, so scripted-case rehearsal and mocks must come from elsewhere.
How many unseen Clinitalk cases or stations should I preserve for final MRCGP SCA calibration? Because your real consultations are effectively always unseen, preserve fresh clinic recordings for the run-in rather than replaying old ones, and keep a handful of the free library cases untouched until the final fortnight — aim for at least four to six unseen calibration encounters.
When should I stop using Clinitalk and move to mixed mocks? When your observable behaviours and guideline currency are stable across several recorded consultations — usually the final two to three weeks — shift to full timed mock circuits, ideally human-observed, to rehearse the twelve-case rhythm Clinitalk does not reproduce.
How should I combine Clinitalk with iatroX without duplicating practice? Use Clinitalk to observe and feed back on your real consulting and iatroX to keep the clinical management inside it current on unseen questions — Clinitalk flags a guideline drift, iatroX confirms and tests the correct plan. This is the two-source design in the two-Q-bank rule, applied across a feedback tool and a knowledge bank; our comparison hub helps you place both.
The bottom line for ST3 trainees
The honest verdict on Clinitalk for the SCA: not a scripted AI-patient simulator but an AI feedback tool on your own real consultations, plus a small free case library and an examiner-led course — and, read that way, a genuinely valuable and distinctive resource. Its strength is exactly what MCQ banks and scripts miss: structured, guideline-referenced feedback on how you actually consult real patients. Its limitation is equally clear: it needs you in clinic, it does not manufacture unseen exam cases at volume, and it does not run a full mock circuit. Use it for the consultation-feedback job it does well, add scripted-case rehearsal and mocks for the rest, calibrate its interpersonal commentary against a human, and let an unseen knowledge source measure whether your management is current. Clinitalk shows you how you consulted; iatroX checks the medicine you consulted with.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 19 July 2026; Clinitalk figures (real-consultation recording and feedback, a free case library of around twenty cases, an examiner-led one-day course, and tiered pricing from roughly £5 to £295) are vendor-reported and were checked on that date — confirm current tiers, allowances, library count and any voice-AI features on the product page. MRCGP SCA format, domains and fee are per the RCGP. Disclosure: iatroX operates a competing clinical-knowledge and question-bank platform and is not a consultation simulator, so this audit names Clinitalk's genuine strengths plainly and confines iatroX's role to the unseen knowledge measurement the product does not claim. Corrections via the feedback route on iatrox.com. References: RCGP Simulated Consultation Assessment pages (rcgp.org.uk); Clinitalk product pages (clinitalk.co.uk); related reading: calibrating automated feedback before you trust the score and why your Q-bank percentage is not your exam score.