More than 1 billion people worldwide are living with a mental health condition (World Health Organization, World Mental Health Today 2025), yet the global median is just 13 mental health workers per 100,000 people. In the US, 69% of adults say they needed more emotional support in the past year than they received (American Psychological Association, Stress in America 2025), and 16% have already turned to an AI chatbot for mental health information or advice (KFF, Tracking Poll on Health Information and Trust 2026). Voice journaling sits right in that gap: a private, spoken habit that AI now transcribes, summarizes and, increasingly, analyzes. We aggregated data from the WHO, SAMHSA’s National Survey on Drug Use and Health, the CDC, the APA, KFF, Pew Research Center, Common Sense Media, NEJM AI, the Annals of Family Medicine, JMIR, OpenAI, Sensor Tower and three market research firms. For the wider app market, see our mental health app statistics roundup.
TL;DR
- More than 1 billion people live with a mental health condition (WHO, World Mental Health Today 2025).
- 23.4% of US adults, about 61.5 million people, had any mental illness in 2024 (SAMHSA, NSDUH 2024).
- Only 14.0% of US adults received counseling or therapy in 2024 (CDC NCHS, NHIS Data Brief 2024).
- 69% of US adults needed more emotional support than they got, up from 65% (APA, Stress in America 2025).
- 16% of US adults asked an AI chatbot about mental health in the past year; 58% of them did not follow up with a professional (KFF, 2026).
- 12% of US teens have gotten emotional support or advice from a chatbot (Pew Research Center, February 2026).
- 0.15% of ChatGPT weekly users show explicit indicators of potential suicidal planning (OpenAI, October 2025).
- Mental health app market: 9.94 billion dollars in 2025, 22.73 billion by 2030 (MarketsandMarkets, 2025).
- Therabot cut depression symptoms by 51% on average in the first generative AI therapy RCT (Dartmouth, NEJM AI 2025).
- A voice biomarker model flagged moderate to severe depression with 71.3% sensitivity and 73.5% specificity (Annals of Family Medicine, 2025).
- Median day-30 retention for popular mental health apps: 3.3% (Journal of Medical Internet Research, 2019).
- 56% of US psychologists used AI at work in 2025, up from 29% in 2024 (APA, Practitioner Pulse Survey 2025).
1. The Unmet Need Behind Every Journaling App
The case for self-guided tools starts with a capacity problem, not a technology trend. More than 1 billion people are living with mental health disorders, and depression and anxiety alone cost the global economy an estimated 1 trillion dollars a year (WHO, September 2025). Against that, governments still spend a median of 2% of their health budgets on mental health, unchanged since 2017, and in low-income countries fewer than 10% of affected people receive care, versus more than 50% in higher-income nations.
The WHO frames this as a workforce gap: with 13 mental health workers per 100,000 people, most of the world cannot be served one appointment at a time. That is the structural reason journaling, meditation and AI conversation apps keep finding an audience, and also the reason regulators worry about them filling the gap with something less than care.
| Metric | Value | Source |
|---|---|---|
| People living with mental health conditions worldwide | More than 1 billion | WHO, World Mental Health Today 2025 |
| Annual global cost of depression and anxiety | About 1 trillion dollars | WHO, 2025 |
| Mental health workers per 100,000 people (global median) | 13 | WHO, World Mental Health Today 2025 |
| Median government health budget spent on mental health | 2% (unchanged since 2017) | WHO, 2025 |
| Share of affected people receiving care: low-income vs higher-income countries | Under 10% vs over 50% | WHO, 2025 |
| Suicide deaths worldwide in 2021 | 727,000 | WHO, 2025 |
The US data shows the gap is not only a low-income-country problem. 23.4% of US adults (about 61.5 million) had any mental illness in 2024, and only 52.1% of them received mental health treatment (SAMHSA, 2024 National Survey on Drug Use and Health, as summarized by NACo). Young adults carry the highest load: 33.2% of 18 to 25 year olds had any mental illness.
Counseling remains a minority experience. The CDC’s National Health Interview Survey data brief found that 14.0% of adults received counseling or therapy from a mental health professional in 2024, ranging from 20.2% of 18 to 29 year olds to 5.0% of adults 75 and older. And the demand for someone to talk to is rising: the APA’s Stress in America 2025 survey (Harris Poll, 3,199 adults, August 2025) found 69% needed more emotional support than they received.
| Metric | Value | Source |
|---|---|---|
| US adults with any mental illness, 2024 | 23.4% (61.5 million) | SAMHSA, NSDUH 2024 |
| Young adults 18-25 with any mental illness | 33.2% | SAMHSA, NSDUH 2024 |
| Adults with any mental illness who received treatment | 52.1% | SAMHSA, NSDUH 2024 |
| Adults with serious mental illness who received treatment | 70.8% (10.3 million) | SAMHSA, NSDUH 2024 |
| US adults reporting generalized anxiety symptoms | 21.7% | SAMHSA, NSDUH 2024 |
| Adults who received counseling or therapy, 2024 | 14.0% (20.2% at 18-29, 5.0% at 75+) | CDC NCHS, NHIS Data Brief 2024 |
| Adults who needed more emotional support than they received | 69% (65% in 2024) | APA, Stress in America 2025 |
| Adults who feel isolated at least some of the time | 54% | APA, Stress in America 2025 |
2. Market Size: Three Forecasts, One Direction
Every firm agrees the mental health app market is growing in the mid to high teens per year; they disagree on the starting point and the slope. MarketsandMarkets sizes the market at 9.94 billion dollars in 2025, rising to 22.73 billion by 2030 at an 18.0% CAGR (MarketsandMarkets, Mental Health Apps Market 2025-2030). Grand View Research is more conservative at 8.87 billion dollars in 2025 and 17.52 billion by 2030 (14.6% CAGR), while Global Market Insights projects 40.9 billion by 2035.
The more revealing numbers are inside the segments. Free apps account for 73.9% of the market and the top five players (Calm, Headspace, Talkspace, Teladoc Health and Woebot) together hold only about 15%, with leader Calm above 5% (Global Market Insights, 2026-2035 forecast). That is a fragmented, freemium-heavy category in which a new AI journaling app can still find room, and in which most users never pay.
| Metric | Value | Source |
|---|---|---|
| Market size 2024 / 2025 | 8.21 / 9.94 billion dollars | MarketsandMarkets, 2025 |
| Forecast 2030 (CAGR 2025-2030) | 22.73 billion dollars (18.0%) | MarketsandMarkets, 2025 |
| Market size 2024 / 2025 / 2030 (CAGR) | 7.48 / 8.87 / 17.52 billion dollars (14.6%) | Grand View Research, Mental Health Apps Market Report |
| Market size 2025 / 2026 / 2035 (CAGR 2026-2035) | 8.2 / 9.6 / 40.9 billion dollars (17.4%) | Global Market Insights, 2026 |
| North America share of the market | 47.0% (2024) vs 57.6% (2025) | MarketsandMarkets vs Global Market Insights |
| Free apps share of revenue | 73.9% | Global Market Insights, 2025 data |
| Depression and anxiety management share | 29.3% | Global Market Insights, 2025 data |
| Top five players’ combined share | About 15% (Calm over 5%) | Global Market Insights, 2025 data |
Context note: the North America share gap (47.0% vs 57.6%) reflects different base years and category definitions, not a real swing. Grand View Research figures were read from its public report summary.
App store data shows where the money is moving. Sensor Tower counted 3.96 billion Health and Fitness app downloads in 2025, up just 0.8%, while in-app purchase revenue rose 13% to a record 4.5 billion dollars, a jump it attributes partly to AI personalization (Sensor Tower, Health and Fitness Apps See Surging Revenue Fueled by AI). Flat installs with rising revenue means the category is monetizing existing users harder, not finding many new ones.
| Metric | Value | Source |
|---|---|---|
| Health and Fitness app downloads, 2025 | 3.96 billion (+0.8% YoY) | Sensor Tower, 2026 |
| Health and Fitness in-app purchase revenue, 2025 | 4.5 billion dollars (+13% YoY) | Sensor Tower, 2026 |
| Health and Fitness apps bidding on AI-themed keywords | 28% (Q4 2025) | Sensor Tower, 2026 |
| Average AI keywords per app | 8 (double Q4 2024) | Sensor Tower, 2026 |
3. AI as a Confidant: Who Is Already Talking to Chatbots
The behavior is already mainstream among younger adults, regardless of what clinicians think of it. 28% of US adults under 30 asked an AI chatbot for mental health information or advice in the past year, versus 8% of adults 50 and older (KFF Poll, March 2026, 1,343 adults, February 24 to March 2, 2026). Across all adults the figure is 16%, and 32% used AI for some kind of health information.
The follow-up number is the one to watch: 58% of people who asked AI about mental health did not then consult a doctor or other professional, compared with 42% for physical health questions. Privacy is part of the appeal, with 36% of AI health users citing it as a reason, which matches what voice journaling users typically say they want: a space to speak without being judged. Our AI companion apps statistics cover the companionship side of this trend in more depth.
| Metric | Value | Source |
|---|---|---|
| US adults who used AI chatbots for health information, past year | 32% | KFF, 2026 |
| US adults who sought mental health information from AI | 16% | KFF, 2026 |
| Under 30 vs 50 and older | 28% vs 8% | KFF, 2026 |
| Did not follow up with a professional after mental health questions | 58% | KFF, 2026 |
| AI health users citing privacy as a reason | 36% | KFF, 2026 |
| AI health users citing that they cannot afford care | 19% | KFF, 2026 |
Teens are further ahead. 64% of US teens use AI chatbots and about three in ten do so daily (Pew Research Center, Teens, Social Media and AI Chatbots 2025), and 12% say they have gotten emotional support or advice from one (Pew Research Center, How Teens Use and View AI, 1,458 teens, September 25 to October 9, 2025). Common Sense Media found that 72% of teens 13 to 17 have used AI companions at least once, and about a third of teen users have chosen to discuss serious matters with an AI instead of a real person (Common Sense Media, July 2025).
At platform scale, emotional use is a small share of a very large number. OpenAI’s own analysis found that only 1.9% of ChatGPT messages concern ‘Relationships and Personal Reflection’ (NBER Working Paper 34255, How People Use ChatGPT, 2025), while non-work use has grown from 53% to more than 70% of messages. OpenAI separately estimated that 0.15% of weekly active users have conversations with explicit indicators of potential suicidal planning and 0.07% show possible signs of psychosis or mania (OpenAI, Strengthening ChatGPT’s responses in sensitive conversations, October 2025). With roughly 800 million weekly users, 0.15% is more than a million people a week.
| Metric | Value | Source |
|---|---|---|
| US teens who use AI chatbots | 64% (about 3 in 10 daily) | Pew Research Center, December 2025 |
| Teens who got emotional support or advice from chatbots | 12% | Pew Research Center, February 2026 |
| Teens who used chatbots for casual conversation | 16% | Pew Research Center, February 2026 |
| Teens who have used AI companions | 72% | Common Sense Media, 2025 |
| ChatGPT messages about relationships and personal reflection | 1.9% | NBER Working Paper 34255, 2025 |
| ChatGPT weekly users with explicit suicidal planning indicators | 0.15% | OpenAI, October 2025 |
| ChatGPT weekly users with possible signs of psychosis or mania | 0.07% | OpenAI, October 2025 |
| Mental health experts consulted; reduction in undesired responses | More than 170; 65-80% | OpenAI, October 2025 |
4. The Evidence: Expressive Writing, Spoken Disclosure and AI Therapy Trials
Journaling apps borrow their credibility from decades of expressive writing research, and the honest summary of that research is ‘small but real.’ Frattaroli’s meta-analysis of 146 randomized studies found a positive, significant average effect of r = 0.075 for experimental disclosure (Psychological Bulletin, 2006), with better outcomes when people disclosed in at least three sessions, for at least 15 minutes, and at home or in private. A meta-analysis of adolescent studies found a similarly small mean effect of g = 0.127 (Clinical Psychology Review, 2015). Most recent available data: these meta-analyses date from 2006 and 2015.
The detail that matters for voice journaling is the mode of disclosure. Esterling and colleagues’ 1994 study, titled ‘Emotional disclosure through writing or speaking modulates latent Epstein-Barr virus antibody titers’ (Journal of Consulting and Clinical Psychology), compared writing with talking into a tape recorder and found comparable biological, mood and cognitive effects, as summarized in Pennebaker’s review in Psychological Science (1997). Speaking is not a lesser form of journaling; on the existing evidence, it is an equivalent one. The open question is whether AI summaries and prompts layered on top add anything, and that has not been tested at scale.
| Metric | Value | Source |
|---|---|---|
| Randomized studies in the largest disclosure meta-analysis | 146 | Frattaroli, Psychological Bulletin 2006 |
| Average effect size of experimental disclosure | r = 0.075 (small, significant) | Frattaroli, 2006 |
| Conditions linked to stronger effects | 3+ sessions, 15+ minutes, private setting | Frattaroli, 2006 |
| Mean effect in adolescents | g = 0.127 | Travagin et al., Clinical Psychology Review 2015 |
| Talking into a recorder vs writing | Comparable biological, mood and cognitive effects | Esterling et al., 1994, via Pennebaker 1997 |
| App-based journaling RCT sample (28 days) | 152 participants; gains only with average or higher baseline self-reflection | Behaviour Change, Writing Yourself Well study |
Generative AI therapy now has its first randomized trial. Dartmouth’s Therabot reduced depression symptoms by 51% on average, anxiety by 31% and eating disorder concerns by 19% versus a waitlist control (Dartmouth, March 2025; published in NEJM AI). Participants used it for about 6 hours on average, roughly equivalent to eight therapy sessions, and rated their therapeutic alliance as comparable to a human professional. The caveats are significant: one trial, 210 participants, a supervised research prototype, and a waitlist rather than an active comparison.
| Metric | Value | Source |
|---|---|---|
| Therabot trial participants | 210 (106 treatment, 104 control) | Heinz et al., NEJM AI 2025 |
| Depression symptom reduction | 51% average | Dartmouth, 2025 |
| Generalized anxiety symptom reduction | 31% average | Dartmouth, 2025 |
| Eating disorder concern reduction | 19% average | Dartmouth, 2025 |
| Effect size for depression vs control | d = 0.845 to 0.903 | Heinz et al., NEJM AI 2025 |
| Average use during the trial | About 6 hours (roughly 8 therapy sessions) | Dartmouth, 2025 |
5. Voice as Data: Transcription, Biomarkers and Spoken Habits
Voice is already the default way many people express emotion digitally. WhatsApp users send an average of 7 billion voice messages every day (TechCrunch reporting WhatsApp’s announcement, March 2022), and WhatsApp explained the habit by saying emotion is more natural in voice than in text. Most recent available data: Meta has not updated this figure since 2022. Platform journaling tools have followed: Apple’s Journal app, launched with iOS 17.2 in December 2023 (Apple Newsroom), added transcripts of recorded audio entries in iOS 18, plus logging of journaling time as mindful minutes.
Funding and usage data for dedicated AI journaling apps is thin and mostly self-reported. Rosebud, an AI journaling startup, raised 6 million dollars in seed funding and reported that users had journaled 500 million words and spent more than 30 million minutes in the app (TechCrunch, June 2025). Fast, accurate speech-to-text is what makes spoken journaling searchable, and our medical transcription statistics show how far that accuracy has come in clinical settings.
| Metric | Value | Source |
|---|---|---|
| WhatsApp voice messages sent per day | 7 billion | WhatsApp via TechCrunch, March 2022 |
| Apple Journal app launch | December 2023 (iOS 17.2) | Apple Newsroom, 2023 |
| Apple Journal audio entry transcripts | Added in iOS 18 (2024) | 9to5Mac, iOS 18 Journal coverage |
| Rosebud seed funding | 6 million dollars | TechCrunch, June 2025 |
| Words journaled on Rosebud | 500 million | TechCrunch, June 2025 |
| Minutes spent on Rosebud | More than 30 million | TechCrunch, June 2025 |
The more consequential, and more sensitive, development is analyzing how someone speaks rather than what they say. A machine learning model trained on more than 14,000 voice samples identified moderate to severe depression with 71.3% sensitivity and 73.5% specificity against PHQ-9 scores, using at least 25 seconds of free speech answering ‘How was your day?’ (Annals of Family Medicine, January 2025; 10,442 samples for training, 4,456 for validation). A 2025 JMIR Mental Health systematic review of speech emotion recognition included 14 studies, 57% of them on depression, 21% on suicide risk and 21% on psychotic disorders.
| Metric | Value | Source |
|---|---|---|
| Voice samples used for depression detection model | More than 14,000 | Annals of Family Medicine, 2025 |
| Training / validation split | 10,442 / 4,456 samples | Annals of Family Medicine, 2025 |
| Minimum free speech per sample | 25 seconds | Annals of Family Medicine, 2025 |
| Sensitivity / specificity for moderate to severe depression | 71.3% / 73.5% | Annals of Family Medicine, 2025 |
| Studies in speech emotion recognition review | 14 (57% depression, 21% suicide risk, 21% psychosis) | JMIR Mental Health, 2025 |
Context note: roughly 7 in 10 accuracy means a voice tool flags many true cases and misses many others. For a voice journal, that supports optional ‘check in with someone’ nudges, not automated diagnosis. The same voice patterns that matter to these models are also why people care about who hears their recordings, a point echoed in our voice assistant statistics on privacy.
6. Retention, Trust and Regulation
Download counts flatter this category; daily use does not. A panel analysis of 93 popular mental health apps found a median of 3.9% of users still opened the app on day 15, and 3.3% on day 30 (Baumel et al., Journal of Medical Internet Research, 2019). Apps with peer support or mindfulness content held users best. Most recent available data: no comparable independent panel study has been published since, so this remains the standard benchmark. For a journaling product, the implication is plain: habit formation is the product, and lowering effort to ‘press and speak’ is a retention strategy.
Clinicians are adopting AI faster than they trust it. 56% of US psychologists used AI tools at work at least once in the past 12 months, up from 29% in 2024, while 92% cited concerns (APA, 2025 Practitioner Pulse Survey; 1,742 respondents, September 2025). Note-taking and dictation was used by 22%, but only 5% used chatbot assistance for clients. Data breaches led the worries at 67%, up from 59%.
| Metric | Value | Source |
|---|---|---|
| Median 15-day / 30-day retention, mental health apps | 3.9% / 3.3% | Journal of Medical Internet Research, 2019 |
| Psychologists who used AI at work, past 12 months | 56% (29% in 2024) | APA, Practitioner Pulse Survey 2025 |
| Psychologists using AI at least monthly | 29% (11% in 2024) | APA, Practitioner Pulse Survey 2025 |
| Psychologists using AI for note-taking and dictation | 22% | APA, Practitioner Pulse Survey 2025 |
| Psychologists using chatbot assistance for clients | 5% | APA, Practitioner Pulse Survey 2025 |
| Psychologists concerned about AI / about data breaches | 92% / 67% | APA, Practitioner Pulse Survey 2025 |
Public opinion is skeptical too. 39% of US adults say chatbots hurt people dealing with loneliness, versus 19% who say they help, and the split is 36% vs 17% for depression and 29% vs 22% for stress (Pew Research Center, August 2026; 3,488 adults, June 2026). Stress, the most journaling-like use case, is where views are closest to even. KFF found 77% of adults concerned about the privacy of health information shared with AI tools.
Regulators are drawing a line between support and treatment. Illinois signed the Wellness and Oversight for Psychological Resources Act (HB 1806) on August 4, 2025, prohibiting AI from making therapeutic decisions while allowing administrative and supplementary uses, with civil penalties of up to 10,000 dollars per violation. The FDA’s Digital Health Advisory Committee met on November 6, 2025 to discuss generative AI-enabled digital mental health medical devices (Federal Register notice). A journaling tool that records and reflects stays on the wellness side of that line; one that diagnoses or treats does not.
| Metric | Value | Source |
|---|---|---|
| Adults saying chatbots hurt vs help people with loneliness | 39% vs 19% | Pew Research Center, 2026 |
| Hurt vs help: depression | 36% vs 17% | Pew Research Center, 2026 |
| Hurt vs help: stress | 29% vs 22% | Pew Research Center, 2026 |
| Adults concerned about privacy of health data shared with AI | 77% | KFF, 2026 |
| Illinois AI therapy law signed | August 4, 2025 (penalty up to 10,000 dollars) | Illinois IDFPR, 2025 |
| FDA advisory committee on generative AI mental health devices | November 6, 2025 | FDA / Federal Register, 2025 |
Summary: AI Voice Journaling and Mental Health Apps by the Numbers
| Metric | Value | Source |
|---|---|---|
| People living with mental health conditions | More than 1 billion | WHO, 2025 |
| Mental health workers per 100,000 people | 13 | WHO, 2025 |
| US adults with any mental illness | 23.4% | SAMHSA, NSDUH 2024 |
| US adults with any mental illness who got treatment | 52.1% | SAMHSA, NSDUH 2024 |
| US adults who received counseling or therapy | 14.0% | CDC NCHS, 2024 |
| US adults needing more emotional support | 69% | APA, Stress in America 2025 |
| Mental health app market 2025 / 2030 | 9.94 / 22.73 billion dollars | MarketsandMarkets, 2025 |
| Free apps share of market | 73.9% | Global Market Insights, 2025 data |
| Health and Fitness in-app revenue 2025 | 4.5 billion dollars (+13%) | Sensor Tower, 2026 |
| US adults asking AI about mental health | 16% (28% under 30) | KFF, 2026 |
| Teens getting emotional support from chatbots | 12% | Pew Research Center, 2026 |
| ChatGPT messages on relationships and reflection | 1.9% | NBER Working Paper 34255, 2025 |
| ChatGPT weekly users with suicidal planning indicators | 0.15% | OpenAI, 2025 |
| Randomized studies in disclosure meta-analysis | 146 (r = 0.075) | Frattaroli, 2006 |
| Talking vs writing disclosure | Comparable effects | Esterling et al., 1994 |
| Therabot depression symptom reduction | 51% | Dartmouth, NEJM AI 2025 |
| Voice biomarker sensitivity / specificity | 71.3% / 73.5% | Annals of Family Medicine, 2025 |
| Mental health app 30-day retention (median) | 3.3% | JMIR, 2019 |
| Psychologists using AI at work | 56% | APA, Practitioner Pulse 2025 |
| Adults saying chatbots hurt vs help with loneliness | 39% vs 19% | Pew Research Center, 2026 |
Methodology and Sources
Every figure was traced to the organization that produced it. Where a primary document could not be opened (several PDFs and paywalled reports), we cite the figure only if it appeared in the publisher’s own summary or in a reporting outlet quoting it, and we name that outlet. Widely shared figures we could not trace to a primary document, including app-reported ‘improvement after 7 days’ claims and generative AI use-case rankings built from online forum posts, were excluded. Nothing here is medical advice.
- World Health Organization: Over a billion people living with mental health conditions (September 2025)
- SAMHSA, 2024 National Survey on Drug Use and Health, via NACo summary, and the 2024 NSDUH mental health treatment report
- CDC NCHS: Mental Health Treatment Among Adults: United States, 2024
- American Psychological Association: Stress in America 2025 and 2025 Practitioner Pulse Survey
- MarketsandMarkets: Mental Health Apps Market 2025-2030; Grand View Research: Mental Health Apps Market Report; Global Market Insights: Mental Health Apps Market 2026-2035
- Sensor Tower: Health and Fitness Apps See Surging Revenue Fueled by AI
- KFF: Poll: 1 in 3 adults are turning to AI chatbots for health information (2026)
- Pew Research Center: How Teens Use and View AI (February 2026), Teens, Social Media and AI Chatbots 2025, Do Americans think chatbots help or hurt (August 2026)
- Common Sense Media: Nearly 3 in 4 teens have used AI companions (July 2025)
- NBER: How People Use ChatGPT, Working Paper 34255 (2025); OpenAI: Strengthening ChatGPT’s responses in sensitive conversations (October 2025)
- Frattaroli: Experimental disclosure and its moderators, Psychological Bulletin (2006); Travagin et al.: expressive writing in adolescents, Clinical Psychology Review (2015); Pennebaker: Writing About Emotional Experiences as a Therapeutic Process, Psychological Science (1997); Writing Yourself Well, Behaviour Change
- Dartmouth: First therapy chatbot trial yields mental health benefits (March 2025, NEJM AI)
- Annals of Family Medicine: Evaluation of an AI-based voice biomarker tool (2025), with details from the EurekAlert release; JMIR Mental Health: Speech emotion recognition systematic review (2025)
- Baumel et al.: Objective User Engagement With Mental Health Apps, JMIR (2019)
- WhatsApp voice messages via TechCrunch (2022); Apple: Journal app launch (2023); Rosebud via TechCrunch (2025)
- Illinois IDFPR: AI therapy law announcement (2025); Federal Register: FDA Digital Health Advisory Committee meeting notice (2025)
- Data watch: Market sizing conflicts substantially: 2025 estimates range from 8.2 billion (Global Market Insights) to 9.94 billion dollars (MarketsandMarkets), and CAGRs from 14.6% (Grand View Research) to 18.0%; none of the firms breaks out voice or AI journaling as a segment, so no reliable ‘voice journaling market size’ exists. Expressive writing meta-analyses (2006, 2015), the Esterling voice disclosure study (1994), the app retention panel (2019) and WhatsApp’s voice message count (2022) are older than three years and shown as most recent available data. The 0.15% suicidal-indicator figure is OpenAI’s own estimate, which OpenAI describes as hard to measure. Therabot results come from one trial against a waitlist control. Rosebud usage figures are company-reported. SAMHSA’s primary NSDUH PDFs were unreachable during research, so NSDUH prevalence figures are cited through the NACo summary and the treatment figures (52.1% and 70.8%) through SAMHSA’s 2024 NSDUH treatment report summary. The Apple Journal transcription feature is documented in iOS 18 coverage such as 9to5Mac, not in an Apple press release. The 2025 NSDUH, expected in mid-2026 to late 2026, will update the US prevalence and treatment figures.
Last updated: October 3, 2026. We update this roundup quarterly, and the next refresh is expected when SAMHSA publishes the 2025 National Survey on Drug Use and Health and Sensor Tower releases its State of Mobile 2027 report.