Blog
Evidence, argument, and the occasional rant — from a working CBT supervisor.
Why Wysa and Woebot stumbled — and what therapist-directed AI does differently
There was a moment around 2023 when autonomous AI mental health chatbots looked like they might replace therapists. That moment has passed. The reasons are structural, not technological — and they shape what useful AI in mental health actually looks like.
Why one platform with four pathways beats four products
The most common reaction from a CBT clinician encountering Supervisia for the first time is to ask which of the four pathways they should use — train, supervise, practise, research. The question contains an assumption worth surfacing: that these are four products with four use cases, and that the user picks the one that fits their current situation. It is a reasonable assumption. It is also the wrong one.
The CBT career arc — train, supervise, practise, research. One platform, every stage.
The CBT career, taken seriously, runs across four distinct phases — and most clinicians are operating in two or three of them at once. The professional infrastructure that supports each phase was built by people who understood that phase and not the others. The cost of that fragmentation is invisible until you try to move data across the boundary.
Building trial-specific training programmes — what 'pivoting on a dime' actually requires
Standard CBT trial training is delivered once, at the start, and treated as a closed step in the protocol. This worked adequately for an earlier generation of trials. It works less well when manuals are mid-revision, when sites are drifting in different directions, and when the trial needs to retrain mid-recruitment. Here is what trial-specific training infrastructure actually has to do, and why repurposing a generic CPD portal does not get you there.
Inter-rater reliability in CBT research — what Cohen's kappa won't tell you alone
Cohen's kappa is the statistic that appears in the methods section of nearly every observational CBT study that reports inter-rater reliability. It is the field's default headline number. It also, on its own, leaves enough about rater calibration undescribed that two studies reporting identical kappa values can have very different actual rater behaviour underneath.
Catching drift in multi-site CBT trials before it contaminates your effect size
In a multi-site CBT trial, the question is not whether some sites will drift. It is which sites and by how much. If you haven't built an instrument to see it, you will read it as treatment failure.
Why CBT trials need continuous fidelity infrastructure, not quarterly tape rating
A trial's reported effect size is a function of two things — the intervention's true efficacy and the fidelity with which it was delivered. The field treats the first as science and the second as logistics. That ordering is backwards, and it quietly degrades a great deal of CBT trial output.
Safeguarding in supervision: why free-text notes aren't enough
The supervisor's role in safeguarding is one of the more consequential aspects of supervisory practice and one of the less robustly instrumented. Most supervision records of safeguarding-relevant conversations live as free-text notes in whatever document the supervisor uses to capture sessions — sometimes structured, often not, retrievable in ways that range from 'in three minutes' to 'if I can remember which laptop I had at the time.'
Co-supervision in DClinPsy training — clinical + course supervisor done right
UK DClinPsy training pairs a placement-based clinical supervisor with a university-based course supervisor. The structure is required by the accrediting bodies and sounds straightforward in description. In practice, the arrangement is asymmetric in ways the formal description does not capture, and most of the operational failure modes live in that asymmetry.
Multi-modal supervision: what BABCP wants you to evidence
BABCP supervisor accreditation has an underappreciated wrinkle — the accreditation is granted per modality, not as a single competence. The supervisor who has accumulated extensive individual supervision experience and applies on that basis is accredited for individual supervision, not for group, peer, or live as their primary supervisory offering. The distinction is made formally and matters at renewal.
Group supervision in NHS Talking Therapies — why it's the norm and how to do it well
Group supervision is, by some distance, the most common form of supervision delivered in NHS Talking Therapies — and one of the least examined. It exists in the workforce because it works under specific constraints, not because anyone thinks it is the optimal mode. Here is what it actually is, what it is good for, where it predictably fails, and what operationally good group supervision looks like.
The supervisor's own CPD — track your own development, not just your supervisees'
The supervisor is, by definition, supposed to be supporting other people's professional development. The slightly inconvenient implication is that the supervisor's own ongoing development also has to keep moving, and BABCP renewal cycles look for supervision-specific CPD as a distinct stream from clinical CPD. Here is what the requirement actually expects, why the standard logging approach fails it, and what good supervisor CPD infrastructure looks like.
Driscoll vs Gibbs vs Rolfe: choosing a reflective practice model that fits how you actually think
Three reflective practice models dominate the UK healthcare landscape: Driscoll's three questions, Gibbs's six-stage cycle, Rolfe's elaborated layered framework. None is best in any general sense. They suit different cognitive styles, and the choice that matters is not which model you pick but whether you use one consistently enough to produce a record by the time anyone asks for one.
The supervision contract is the most ignored part of supervision (and why that's a problem)
Almost every CBT supervisor knows they should have a written supervision agreement. Almost every supervisor believes they do. The actual contract — the one that would survive a complaint, an accreditation review, or a fitness-to-practise question — usually does not exist in usable form. Here is what the standards ask for, why a one-page template will not get you there, and what a live contract has to do.
What BABCP supervisor accreditation actually needs — and how to assemble it without a weekend lost
BABCP supervisor accreditation is a reasonable credential built on reasonable requirements. The labour cost lives almost entirely in evidence retrieval — years of supervision practice that was never organised for portfolio assembly. Here's what the criteria ask for, what the recurring failure mode looks like, and what good portfolio infrastructure does differently.
Catching alliance rupture before dropout — what the evidence actually supports
Around one in five adult psychotherapy clients drops out before agreed termination, and weak alliance is the most consistent predictor. The Safran and Muran rupture framework gives the field a language for what is happening; the Eubanks meta-analysis shows that repair, when it happens, improves outcomes. The remaining problem is detection — and it is mostly an instrumentation problem.
Why session-by-session ROM beats intake-and-discharge measurement
Most outcome measurement in routine CBT is intake-and-discharge — a PHQ-9 at the start, a PHQ-9 at the end, and a subtraction in between. That is measurement in the same sense a thermometer at the start and end of an illness is measurement: technically true, of no use to anyone trying to treat the patient in between.
The CBT homework problem — what the Kazantzis meta-analysis actually says
Between-session work is one of the most replicable predictors of CBT outcome, and one of the most quietly neglected ingredients of routine practice. Here's what the Kazantzis meta-analysis really shows — and why the bottleneck is almost always structural.
Why AI clients beat role-play with classmates
Classroom role-play is the default training infrastructure for CBT skill development, and it is doing some useful work. It is also doing less than the curriculum tends to claim. Here is what role-play actually delivers, where it hits a ceiling, and why AI clients sit closer to the kind of practice the deliberate-practice and fidelity literatures point to.
CTS-R: what it is, how it's scored, and why it should be on every CBT therapist's desk
The Cognitive Therapy Scale — Revised is the closest the field has to a working language for observable CBT competence. Most qualified therapists encountered it once in training and quietly retired it. That is a mistake, and the reasons are structural rather than sentimental.
Deliberate practice — the missing layer between qualification and expertise
Qualification proves you've learned the model. It doesn't prove you can deliver it. What 30 years of expertise research says about the difference — and what it means for any clinician who finished training and assumed the rest would follow.
Why therapist drift is the dirty secret of CBT outcomes
Qualified CBT therapists routinely move away from the techniques they were trained to deliver. The literature has been telling us this for fifteen years. Here's what the evidence actually says — and what it implies for your practice.
The History of CBT: From Behaviourism to the Third Wave
From behaviour therapy to the cognitive revolution to mindfulness — trace the evolution of the most researched psychotherapy in history.
Beck's Cognitive Triad: Understanding Depression
How negative views of self, world, and future maintain depression — and what CBT does about it.
Behavioural Activation: When and Why It Works
Action before motivation — the evidence-based behavioural approach to depression that rivals full CBT.
Intolerance of Uncertainty: The Heart of GAD
Why people with Generalised Anxiety Disorder worry endlessly — and how to treat the engine, not just the symptoms.
Why We Formulate: The Bridge Between Theory and Practice
Formulation vs. diagnosis, shared understanding, and why every intervention should be formulation-driven.