Psychodynamic therapy vs CBT: what actually differs

What each approach does

Both are talking therapies, delivered by a trained clinician, and both have been tested in randomised trials. They differ in what they treat as the problem and in how a session is run.

The CBT column follows the NHS description: a practical therapy that examines how difficult situations affect thoughts, feelings and behaviour, with skills practised between sessions.

The psychodynamic column follows Jonathan Shedler's 2010 review, which draws on process studies that coded real session recordings. It lists seven features that reliably separated psychodynamic technique from CBT: a focus on emotion, exploration of what the person avoids, identification of recurring themes and patterns, discussion of past experience, attention to relationships, attention to the relationship with the therapist, and free speech about wishes and fantasies. The last point is the sharpest contrast: where CBT usually follows an agenda, psychodynamic work follows the person.

What the evidence shows

Three findings are well supported.

CBT has the broader evidence base. A 2012 review identified 269 meta-analyses of CBT and examined a sample of 106. Support was strongest for anxiety disorders, somatoform disorders, bulimia, anger control and general stress. The authors called the evidence strong, and noted that many subgroups were under-studied.

Psychodynamic therapy also works, on average, in adults. Shedler's review reports large effects for short-term psychodynamic therapy against control conditions, and argues that gains persist or grow after treatment ends. That is one author's reading of the data, in a narrative review that argues a position. A larger effect at follow-up has more than one possible explanation.

Head-to-head, adults improve by similar amounts. A 2017 meta-analysis pooled 23 randomised trials with 2,751 patients, and ran a formal equivalence test against a pre-set margin of 0.25 standard deviations. For target symptoms, the confidence intervals sat inside that margin at the end of treatment and at follow-up. Most comparisons were against CBT, per the abstract. A 2021 network meta-analysis of 331 depression trials (34,285 patients) reached a similar picture: the main psychotherapies did not differ significantly from each other, and the authors suggested patient preference and availability may matter more when choosing.

What "no significant difference" does and does not mean

Reading these results correctly takes care, because an absence of evidence of a difference is not evidence that there is none.

Psychotherapy trials are usually too small to detect a modest gap. One analysis of more than 100 trials in adult depression calculated that finding a difference of 0.24 standard deviations would need about 548 patients per trial. The largest trial comparing CBT with another therapy had 178 patients, enough to detect only a difference of 0.42 or more. The author's conclusion was that individual trials are heavily underpowered, so it remains unclear whether clinically relevant differences exist.

That is why the 2017 equivalence analysis matters: it tests for similarity directly instead of inferring it from a failed test for difference. It has limits too. The margin of 0.25 is a judgement about how large a gap would matter. It covers target symptoms, not everything a person might want from therapy. And the pooled trials differ in disorder, length and therapist.

Two further cautions follow from the same literature.

• Averages hide the individual. Equal average benefit does not say which approach suits you. The 2021 depression analysis found psychodynamic therapy better than care as usual and waiting lists, but not better than a pill placebo. That is a finding about depression specifically, and it means this evidence should not be stretched to every problem.
• Age changes the picture. In a 2015 network meta-analysis of 52 studies in young people with depression (3,805 participants), interpersonal therapy and CBT outperformed most controls at the end of treatment, while psychodynamic therapy was not significantly better than a waiting list. Fewer trials exist for the psychodynamic approach, so this shows thin evidence, not proof of no benefit.

How the difference shows up in a real life

Take a type of situation, not a case: a person who understands precisely why they avoid conflict, and goes silent every time it comes close.

A CBT clinician would be likely to treat the silence as a behaviour to change. They would map the thought that comes before it ("if I say this, they will leave"), test that thought, and set graded practice between sessions. A psychodynamic clinician would be likely to treat the silence as information: when did it start, who did it protect you from, does it happen in this room too. The first aims at a different response next week. The second aims at understanding why the pattern exists, in the expectation that it loosens as that becomes clear.

Neither description is a rule. The 2010 review suggests that some non-psychodynamic therapies may work partly because skilled practitioners use techniques central to psychodynamic theory, so the boundary is blurrier in a consulting room than on a page.

How to choose

• If the problem is specific and bounded (a phobia, panic, a compulsion, an eating problem), start with a clinician who offers a structured, skills-based treatment. The CBT evidence base is widest there.
• If the problem is a pattern that keeps returning across relationships or jobs, and you want to understand it rather than only manage it, psychodynamic work is designed for that question.
• If you are unsure, ask a prospective therapist how they would work with your problem, how long they expect it to take, and how they will know whether it is helping. A good answer to those three questions tells you more than the label.
• Fit with the person counts. The 2021 analysis found few differences between therapies, which puts weight on whether you can work with this clinician.

Where TwentyThird fits

TwentyThird sits on the psychodynamic side of this comparison. It is a psychodynamic AI platform for self-understanding: a member writes with an AI agent named Oneli, which tracks what repeats in what they say, between appointments. A licensed psychologist configures and supervises that agent. Psychologist supervision is in early access, not every member has one yet, and no human is alerted today if a member writes something alarming.

TwentyThird is not therapy and does not claim to deliver either approach. None of the studies above tested TwentyThird or any AI tool, so they say nothing about whether it helps. You can read how it works on the methodology page and the research it draws on on the science page.

What this does not replace

This article is general information, not clinical advice, and it cannot tell you which therapy fits your situation. A licensed psychologist, psychotherapist or physician can assess that, and can tell you when a structured treatment is the right first step. Neither therapy is the right answer to an emergency. If you are in danger, contact local emergency services now. In the EU, call 112.