Does Therapy Really Work? What Decades of Research Say

"Does therapy actually work?" It's a fair question — and one we're asked often, usually by someone who is on the fence about starting. The honest answer, backed by decades of research spanning hundreds of thousands of patients, is a well-evidenced yes. But the more useful answer lies in the specifics: how much it helps, how long it takes, and for whom.

Where the evidence comes from

Psychotherapy is one of the most extensively studied interventions in medicine. Since the first major meta-analysis in 1980 combined nearly 500 individual studies, researchers have continued pooling results across thousands of trials. A large 2022 umbrella review published in World Psychiatry, which combined the findings of the largest meta-analyses across every major mental health condition, found that psychotherapy produces a standardised effect size of approximately 0.34 — statistically on par with the effect size found for medication (0.36) across the same conditions.

In routine clinical practice — not just controlled trials — a 2010 review pooling 31 studies found psychotherapy to be highly effective for common mental health conditions, with an even larger effect size than typically found in stricter research settings. In other words: therapy doesn't just work in idealised laboratory conditions. It works in the kind of ordinary clinical setting most people actually walk into.

How many sessions does it actually take?

This is usually the real question underneath "does it work" — and it has a surprisingly well-documented answer. Since the 1980s, researchers have tracked what's called the dose-response relationship in psychotherapy: essentially, how much benefit accumulates as the number of sessions increases.

Based on the landmark Howard et al. dose-response curve, replicated across multiple subsequent studies.

The pattern that emerges consistently across studies is what researchers call a "negatively accelerating curve" — most of the change happens earlier than people expect, and the general consensus across the research is that somewhere between 13 and 18 sessions is the point at which about half of patients reach meaningful improvement. That's roughly three to four months of weekly sessions — far shorter than the open-ended, years-long commitment many people picture when they imagine starting therapy.

Does the type of therapy matter?

Yes — but perhaps less than people assume. Across conditions, well-established approaches like Cognitive Behavioural Therapy (CBT) consistently perform strongly in head-to-head comparisons, particularly for anxiety disorders, OCD, and depression. But research also shows that combining therapy with medication, where clinically appropriate, tends to outperform either approach alone — an important consideration for more complex or long-standing conditions rather than a sign that therapy alone has failed.

What the research doesn't capture

Meta-analyses are powerful, but they average across thousands of different people, which can obscure an important truth: therapy is not one uniform intervention. The fit between a specific client and a specific therapist — sometimes called the "therapeutic alliance" — is itself one of the most consistently replicated predictors of good outcomes, independent of which specific technique is used. This is part of why we encourage people to view the first session or two as a chance to assess fit, not as a binding commitment.

What this means if you're considering starting

• You are not signing up for years of open-ended commitment — most meaningful change research points to happens within roughly 3-4 months of consistent sessions
• The research strongly supports evidence-based approaches like CBT, particularly for anxiety, OCD, and depression
• How well you connect with your specific therapist matters as much as the technique itself — it's completely reasonable to try a session or two before deciding
• For more complex or longstanding difficulties, therapy combined with medication (where appropriate) tends to outperform either alone