Evidence Review — 2026

The Evidence on Journalling
& Mental Health

What does the research actually say about writing as a mental health intervention? A careful look at the meta-analyses, effect sizes, and honest uncertainties.

At a Glance

Journalling works, but the effects are small and the type matters enormously

The research base spans 40+ years and hundreds of trials. Expressive writing (the Pennebaker paradigm) produces small but statistically significant effects. Gratitude journalling modestly improves well-being. Clinician-guided Written Exposure Therapy is genuinely competitive with gold-standard PTSD treatments. The key variables are structure, clinical guidance, and targeting a specific condition.

d = 0.16
Expressive Writing
9%
Anxiety Reduction
5-8%
Gratitude Effects
100+
Studies Since 1986
6-24%
WET Dropout Rate
g = 0.43
Mental Health Apps
Results have been difficult to replicate and it is unclear what conditions are necessary for observing the effect.
— James Pennebaker, 2018, on his own paradigm

How to read this page

Every finding is tagged with a confidence level based on the quality and consistency of the evidence. Effect sizes are reported as Cohen's d or Hedges' g wherever available. We distinguish between types of journalling — the evidence for structured clinical protocols (Written Exposure Therapy) is far stronger than for generic "writing in a diary." Use the navigation above to explore evidence, compare methods, read fact-checked answers to key questions, or browse what the rationalist community has written about journalling.

Expressive writing has a real but small effect

High Confidence
d = 0.08 – 0.16

Multiple meta-analyses over 100+ studies confirm a small positive effect on mental health. Roughly equivalent to the difference between a coin flip and a 54/46 split. The effect is real but significantly smaller than initially believed (Smyth's 1998 d = 0.47 has not been replicated).

Effects are larger for anxiety and PTSD than depression

High Confidence
9% anxiety | 6% PTSD | 2% depression

Consistent across meta-analyses: journalling helps most with anxiety (9% reduction) and PTSD (6%), less with depression (2%). For depression, interventions longer than 30 days do better (10.4% improvement). Sohal et al. 2022, 20 RCTs.

Written Exposure Therapy is an effective PTSD treatment

High Confidence
Non-inferior to CPT/PE

17 studies including 7 RCTs show WET is as effective as Cognitive Processing Therapy and Prolonged Exposure (gold-standard PTSD treatments) with far fewer sessions (5 vs 12+) and dramatically lower dropout (6-24% vs 36-45%). VA/DoD-recommended. Now proven via telehealth and in group formats.

Gratitude journalling modestly improves well-being

Moderate Confidence
g = 0.19 overall

A 2025 PNAS meta-analysis of 145 studies across 28 countries found overall g = 0.19. Effects on happiness (g = 0.25) and well-being (g = 0.30) are larger than for depression (g = 0.13). Effectiveness varies by country — works in the US, China, Germany; not in France, Japan, UK. Evidence quality often rated "very low" to "low."

Journalling can make things worse through rumination

Moderate Confidence

Abstract, evaluative writing ("Why did this happen to me?") can increase intrusive memories and worsen symptoms. Writing focused only on emotions without sense-making led to worse illness symptoms (Ullrich & Lutgendorf, 2002). A meta-analysis of 45,287 adolescents confirms rumination significantly harms well-being across all mental health domains.

Clinical Evidence

Meta-Analyses and Key Findings

Every major meta-analysis on journalling and mental health, with effect sizes and confidence levels.

Smyth (1998) — Written Emotional Expression

Meta-analysis Moderate
d = 0.47

First major meta-analysis. Found health enhanced across 4 outcome types: reported physical health, psychological well-being, physiological functioning, and general functioning. Health behaviours were not influenced. This large effect size has not been replicated in later, more rigorous analyses.

Smyth, J.M. (1998). Journal of Consulting and Clinical Psychology.

Frattaroli (2006) — Experimental Disclosure

Meta-analysis Moderate
r = .075

146 studies. Overall effect size small but significant across physical, psychological, and general functioning outcomes. Large heterogeneity — effects varied dramatically by study.

Frattaroli (2006). Psychological Bulletin.

Sohal et al. (2022) — Journalling for Mental Illness

Meta-analysis, 20 RCTs Moderate
68% of outcomes effective

Anxiety: 9% reduction (highest). PTSD: 6%. Depression: 2% (lowest, but >30 day interventions improved 10.4%). Critical finding: privacy matters — studies collecting journals showed worse outcomes. B-level recommendation.

PMC8935176

Lai et al. (2023) — Expressive vs Positive Writing

Meta-analysis, 24 RCTs Moderate
MD = 0.41 (positive emotions)

Positive writing generated more positive emotions and reduced negative emotions more than expressive writing. But in patients, expressive writing stimulated stronger cognitive processing; in general population, positive writing promoted cognitive mechanisms.

PMC10415981

Reinhold et al. (2018) — Expressive Writing and Depression

Meta-analysis Moderate
Not significant overall

Expressive writing did not yield significant long-term effects on depressive symptoms. Effects were larger with more sessions and more specific topics. Female-majority samples benefited more.

Reinhold et al. 2018

Guo (2023) — Long-term Follow-ups

Meta-analysis Moderate
g = -0.12

Found a delayed, durable effect of expressive writing on depression, anxiety, and stress. Small but significant. Key insight: benefits may appear weeks to months after writing, not immediately.

Guo 2023

PNAS (2025) — Cross-Cultural Gratitude Meta-Analysis

145 studies, 28 countries Moderate
g = 0.19 overall

24,804 participants. Happiness g = 0.25, well-being g = 0.30, depression g = 0.13, negative affect g = 0.05. Effectiveness varies significantly between countries — works in US, China, Germany, Canada, Australia; not in France, India, Japan, Netherlands, UK.

PMC12280877

Written Exposure Therapy — State of the Science (2025)

Clinical protocol High
Non-inferior to CPT/PE; dropout 6-24%

17 published studies including 7 RCTs. Five 50-minute sessions. VA/DoD-recommended for PTSD. Dramatically lower dropout than gold-standard treatments (6-24% vs 36-45%). Now demonstrated via telehealth (d = 1.08), in group formats (78% attendance), and with diverse populations including veterans, adolescents, and Black perinatal women.

VA/DoD | 2025 Review

Self-Compassion Writing Interventions

56 RCTs Moderate
g = 0.51 for self-criticism

Medium effects on self-criticism (g = 0.51), with small-medium effects on depression and anxiety. Brief online self-compassion writing produced medium-to-large reductions in shame, external shame, self-criticism, and general anxiety. Strongest effects for high-baseline self-criticism.

PMC9992917

Mental Health Apps Overall (2025)

92-RCT Meta-analysis Moderate-High
g = 0.43

16,728 participants across 92 RCTs. Apps significantly improved clinical outcomes vs controls with moderate effect size. Surprisingly, more persuasive design features did not predict better outcomes or engagement.

Nature Digital Medicine

Therabot RCT — First AI Therapy Chatbot Trial (2025)

Single RCT Low-Moderate
51% MDD reduction | 31% GAD reduction

First RCT of a generative AI therapy chatbot (NEJM AI). Effect sizes exceeded typical SSRI trials and approached first-line psychotherapy. Therabot incorporates reflective writing prompts and cognitive restructuring. Promising but single study with waitlist control.

NEJM AI 2025

EW for Grief (2025 Meta-Analysis)

13 RCTs Moderate
g = 0.39 (grief) | 0.68 with therapist feedback

Expressive writing works for grief (g = 0.39) and co-occurring depression (g = 0.31). Dose-response confirmed: more sessions improve outcomes (g = 0.58). Therapist feedback nearly doubles the effect (g = 0.68 vs small without).

EW for Postpartum Depression

Meta-analysis, 5 studies Moderate
SMD = -0.65 (depression) | -0.80 (stress)

483 participants. Notably larger effects than general population EW studies. Expressive writing appears particularly effective for postpartum women, possibly because of the high emotional intensity and limited access to other support.

Lim et al. 2025

Effect Size Comparison

How Journalling Compares

Intervention Effect Size (d) Access Cost
CBT for depression0.24 – 0.80Therapist requiredHigh
Antidepressants0.30 – 0.50PrescriptionModerate
Exercise (aerobic)0.50 – 0.70Self-directedLow
Mindfulness meditation0.22 – 0.38Self-directedLow
Expressive writing0.08 – 0.16Self-directedFree
Gratitude journalling0.10 – 0.20Self-directedFree
Written Exposure TherapyNon-inferior to CPT/PEClinician-guidedModerate
Self-compassion writingg = 0.51 (self-criticism)Self-directedFree
Mental health appsg = 0.43App-basedFree–Low
Methods Compared

Different Approaches, Different Evidence

Not all journalling is the same. The type matters enormously for both outcomes and evidence quality.

Expressive Writing (Pennebaker)

Moderate evidence

Write about your deepest thoughts and feelings for 15-20 minutes, 3-4 consecutive days. The most studied journalling intervention. 100+ studies, d = 0.08-0.16. Works best for anxiety and PTSD, less for depression.

Best for: Processing unresolved emotional experiences, trauma, major life transitions.

Gratitude Journalling

Moderate evidence

List 3-5 things you're grateful for. Weekly may work better than daily (avoids gratitude fatigue). ~7% improvements in anxiety, depression, and life satisfaction. Effects vary by country.

Best for: General well-being, mild-moderate low mood. Weekly practice.

Written Exposure Therapy

Strong evidence

5 sessions of structured writing about trauma with clinician guidance. VA/DoD-recommended for PTSD. Non-inferior to gold-standard treatments with far fewer sessions and dramatically lower dropout.

Best for: PTSD. Requires a clinician. Now available via telehealth.

Morning Pages (Cameron)

Weak evidence

Three pages of longhand stream-of-consciousness writing, first thing in the morning. No studies have tested Morning Pages as Cameron designed them. Benefits supported mostly by anecdotal reports.

Best for: Creative professionals, mental clarity. More "mental hygiene" than clinical intervention.

Self-Compassionate Letter Writing

Moderate evidence

Write a letter to yourself from a compassionate perspective. Medium-to-large reductions in shame, self-criticism, and general anxiety. g = 0.51 for self-criticism across 56 RCTs. Strongest effects for high-baseline self-criticism.

Best for: Shame, self-criticism, perfectionism, harsh self-judgement.

Bullet Journalling

Very weak evidence

Customisable organisational system using rapid logging. No RCTs. Primarily a productivity tool. Anecdotally popular in ADHD communities. A 2025 exploratory study identified four mental health usage patterns.

Best for: Organisation, ADHD management, combining planning with reflection.

CBT Thought Records

Strong (as part of CBT)

Identify situation, automatic thoughts, emotions, cognitive distortions, generate alternatives. Integral part of CBT (d = 0.5-0.8 for depression). RNT-specific CBT achieves g = -0.99 for rumination.

Best for: Depression, anxiety. Best used with therapist guidance.

Positive Affect Journalling

Limited but promising

Writing about positive experiences, hopes, and goals. One RCT showed reduced anxiety in medical patients. Positive writing generates more positive emotions than expressive writing (MD = 0.41). May be better for general populations.

Best for: Those who don't want to focus on negatives. Complement to expressive writing.

AI-Guided Journalling Apps

Emerging evidence

AI-guided journalling increased emotional clarity 34% vs unguided (8 weeks). GenAI CBT app increased engagement 2.4-3.8x vs workbooks. AI feedback on writing reduced anxiety more than writing alone. Only 16% of LLM studies have clinical efficacy testing.

Best for: Engagement. Best as a feedback layer on human writing, not a replacement.
Fact-Checked

Key Questions About Journalling

Each question assessed with evidence, a verdict, and a confidence rating.

1 Does journalling actually help with depression and anxiety?

For anxiety: Meta-analysis of 20 RCTs found 9% reduction vs control. Gratitude interventions: 7.8% reduction in GAD-7 scores. Positive affect journalling significantly reduced anxiety in medical patients.

For depression: Only 2% reduction overall, but interventions >30 days showed 10.4% improvement. Gratitude: 6.9% lower PHQ-9. For comparison: antidepressants d = 0.30-0.50; CBT d = 0.50-0.80.

Special populations: Postpartum women show much larger effects (SMD = -0.65 for depression). EW in drug rehabilitation showed significant reductions vs active control.

Verdict: Partially true. Meaningful (though modest) effects on anxiety. For depression, effects are smaller and less consistent. Not a replacement for evidence-based treatments but may help as an adjunct, particularly with sustained practice (>30 days).
Moderate Confidence
2 Is expressive writing (Pennebaker paradigm) robustly replicated?

100+ studies since 1986, with an overall average d = 0.16. Pennebaker himself acknowledged in 2018 that "results have been difficult to replicate." A strict meta-analysis found neither somatic nor psychological variables reached significance. More permissive analyses find small but significant effects (r = .075).

High heterogeneity (I² = 83.8%). Three meta-analyses support benefits; three yield null effects. The field itself is split.

Verdict: Partially true. The effect is "robust" in the sense that something is consistently happening, but not in the sense that any given study reliably produces significant results. Common for small-effect-size psychological interventions.
High Confidence
3 Gratitude journalling — does it actually work?

2025 PNAS meta-analysis (145 studies, 28 countries): overall g = 0.19. Happiness g = 0.25, well-being g = 0.30, depression g = 0.13. Weekly better than daily (avoids fatigue). Effects vary by country — works in US, China, Germany; not in France, India, Japan, Netherlands, UK.

Evidence quality: mostly "very low" to "low" certainty due to blinding issues, high dropout, and measurement inconsistency.

Verdict: Mostly true, with caveats. Improves well-being and modestly reduces anxiety and depression. Effects are small (5-8%), evidence quality often low. A reasonable wellness practice, not a therapy.
Moderate Confidence
4 Is handwriting better than typing for journalling?

fMRI and EEG studies show handwriting activates broader brain networks. A 2025 review confirmed broader functional connectivity vs typing. Handwriting produces theta band synchronisation linked to memory consolidation.

However: No RCTs directly compare handwritten vs typed journalling for mental health outcomes. Brain activation differences in learning studies may not translate to therapeutic writing.

Verdict: Plausible but unproven for mental health. Direct clinical trials don't exist. The medium probably matters less than consistency and engagement.
Low Confidence
5 Can journalling make things worse?

Rumination risk: Abstract, evaluative writing ("Why did this happen to me?") can increase intrusive memories. A meta-analysis of 45,287 adolescents confirmed rumination significantly harms well-being.

Emotion-only focus: Writing focused solely on emotions without cognitive processing led to worse illness symptoms.

Privacy matters: Studies collecting journals showed worse outcomes.

AI safety gaps: Rosebud's CARE benchmark: 24/25 AI models failed at least one critical self-harm safety test.

Verdict: True. Journalling can harm when: (a) abstract rumination replaces concrete processing, (b) writing focuses only on emotions without sense-making, (c) severe trauma without clinical support, or (d) privacy is compromised.
Moderate-High Confidence
6 How does journalling compare to other interventions?

Unguided expressive writing and gratitude journalling have effects roughly 1/3 to 1/5 the size of established treatments. However, clinician-guided Written Exposure Therapy achieves effects comparable to gold-standard PTSD therapies.

Verdict: Mostly false for generic journalling; true for structured clinical protocols. Journalling is a low-cost, low-risk complement, not a replacement. Its advantage is accessibility: free, private, available to anyone. For PTSD specifically, structured writing therapy is genuinely competitive.
High Confidence
7 What's the minimum effective dose?

Acute processing: Pennebaker protocol: 15-20 minutes, 3-4 days. WET: 5 sessions of 30-minute writing.

Ongoing maintenance: Interventions >30 days significantly outperformed shorter ones. Gratitude: weekly better than daily. Consistency matters more than session length.

Dose-response: Grief meta-analysis: more sessions improve outcomes (g = 0.58). Therapist feedback nearly doubles the effect (g = 0.68).

Verdict: Partially answered. For acute processing, 3-4 sessions of 15-20 minutes appears sufficient. For maintenance, >30 days of regular practice. No precise "minimum effective dose" established.
Moderate Confidence
8 Do rationalists and LessWrongers find journalling useful?

The rationalist community engages substantially with the broader benefits of writing — writing to clarify thinking, blogging as epistemic practice. Less engagement with journalling as a mental health intervention specifically.

Key themes: Writing as thinking (Yudkowsky, Paul Graham). Epistemic benefits: writing forces you to notice confusion. The quantified self movement overlaps with journalling. CFAR techniques include structured reflection. Inkhaven writing residency produced extensive community discussion.

Verdict: Partially true. Broadly positive about writing as a thinking tool, but no consensus on journalling as mental health intervention. The community tends to weight modest effect sizes against time investment. The strongest rationalist case is for journalling as a thinking tool, not a therapy.
Moderate Confidence
Community Writing

Rationalist & LessWrong Takes on Journalling

What the rationalist community has written about journalling, writing as thinking, and related practices. Selected highlights from a comprehensive search of LessWrong, SlateStarCodex/ACX, EA Forum, and rationalist blogs.

The rationalist case for journalling is less about therapy and more about epistemic hygiene — writing forces you to notice confusion, identify gaps in understanding, and make implicit beliefs explicit.
— Synthesis from community posts

Journalling Specifically

Peter WildefordJuly 201338LessWrong

Comprehensive evidence review arguing gratitude journalling is one of the most scientifically validated happiness interventions. Weekly journalling produced stronger gains than daily (avoids boredom). Imagining a world without positive elements amplified gratitude's benefits.

ProofOfLogicLessWrong

Rationalist-specific gratitude variant combining it with value affirmation. Writing about what you valued over the past week acts as a weekly review for self-debugging. Value affirmation shows effects a year later after just 15 minutes.

RobertLumleyFebruary 201212LessWrong

Foundational LessWrong post on journalling as a tool for memory preservation, tracking belief changes over time, and facilitating reflective thinking.

ChristianKlLessWrong

Crowdsources community journalling techniques including sentence completion, past/future-self conversations, and ancestor thought experiments.

Peter Hurford~2013Everyday Utilitarian
Everyday Utilitarian: Gratitude Journal

Documents rational discontinuation of gratitude journalling after finding diminishing marginal returns for already-happy individuals. An honest data point: the practice may have a ceiling effect.

Neel NandaLessWrong

Makes the case for weekly reviews as the meta-habit that debugs all other habits. Practical implementation guidance for structured reflection as a keystone practice.

Ben Kuhnbenkuhn.net

Documents a 3-4 hour Saturday review practice that produced 50% focused-work-time improvement through compounding small optimisations.

Writing as Thinking

Holden KarnofskyCold Takes

Former Open Philanthropy CEO on writing as his primary tool for learning and thinking. Writing forces you to confront gaps in understanding you can paper over in conversation.

Jacob FalkovichPutanumonit

Documents Putanumonit's 5-year writing evolution. Claims writer's block is fake. Advocates binge-writing over daily practice.

Gwern2024gwern.net

Gwern's case for public writing as an epistemic practice. Writing online creates a long-term record, forces clarity, invites feedback, and compounds over time.

Zvi Mowshowitz2025thezvi.substack.com

Documents a deliberate shift from output-maximisation to thought-quality-maximisation. Reducing volume to improve quality — a counterpoint to "just write more."

lumenwritesLessWrong

Discussion of thinking-through-writing as methodology. Writing is not just expressing existing thoughts but generating new ones.

CBT, Therapy, and Structured Writing

fortyeridaniaLessWrong

Examines whether CBT thought records constitute epistemic manipulation. Concludes the practice is both instrumentally and likely epistemically rational.

aaqLessWrong

First-person account of a rationalist overcoming intellectual resistance to CBT writing exercises. Adopted ABC worksheets despite philosophical concerns about depressive realism.

Lynette ByeEA Forum
Rewriting My Mindset

EA community member on using CBT techniques and writing to address perfectionism. Practical account of therapeutic writing in the rationalist/EA context.

JonahSLessWrong
Methods for Treating Depression

Surveys treatment approaches with CBT highlighted for its combination of efficacy, lasting changes, low cost, and absence of adverse side effects. Recommends CBT self-help workbooks.

AI, Writing, and Cognitive Outsourcing

Various authors2024-2025LessWrong
Don't Let LLMs Write For You (multiple posts)

Recurring community concern: if the cognitive benefits of writing come from the thinking process, then letting AI write for you eliminates the benefit. AI should be a feedback layer or thinking partner, not a replacement for the act of writing.

MindScape study2024CHI Conference
MindScape: Contextual AI Journaling

LLM + behavioural sensing app reduced negative affect by 11% and improved self-reflection by 6% over 8 weeks. Context-aware AI-generated prompts. Published at CHI.

Various authors2025-2026LessWrong
AI Therapy Risks & "Therapist in the Weights"

Community discussion of risks: AI sycophancy reinforcing maladaptive beliefs, hyper-introspection without resolution, parasocial attachment. The rationalist community is notably more cautious about AI therapy than the general public.

EA Mental Health Infrastructure

EA Forum
The Case for Bibliotherapy

Argues self-help books (particularly CBT workbooks like Feeling Good) match therapist-delivered CBT in effectiveness. Relevant to journalling as CBT workbooks are structured writing exercises.

EA Forum
Kaya Guides — Tech-Enabled Mental Health

EA-backed initiative for 5-10x cheaper psychotherapy through WhatsApp-based guided self-help. Includes structured reflection and writing exercises. Pilot results with WHO Step-by-Step programme.

Rethink WellbeingEA Forum
Deliver Mental Health Outcomes at Scale

Peer-led CBT programme at universities. Includes structured journalling and thought record exercises as core components. Demonstrates scalable approach to mental health using writing-based techniques.

Journalling is a low-cost, low-risk complement to other interventions, not a replacement. Its real advantage is accessibility — it's free, private, and available to anyone who can write.

ai gen