AI Therapist Statistics: Who Uses AI for Therapy in 2026

How many people use AI for therapy

In a 2026 YouGov survey of US adult citizens fielded April 1-5, 6 percent said they had used AI for therapy, and another 17 percent said they had not but would consider it. Estimates depend heavily on how the question is asked, so it helps to separate narrow therapy use from broader mental-health support. When people are asked specifically about therapy, the numbers stay modest.

Broader mental-health and emotional support is more common than narrow therapy use, and use rises sharply among people who already rely on AI. A survey conducted in February 2025 among 499 US adults with ongoing mental-health conditions who had used large language models in the past year, published in the APA journal Practice Innovations, found that 48.7 percent of this selected sample said they were using large language models for therapeutic support.

These figures are not directly comparable: the YouGov number is the share of all adults, while the Sentio number is the share of a narrower group of existing AI users who also have mental-health challenges. Taken together, they suggest that dedicated AI therapy use across the whole population is still a minority behavior, but that it is meaningfully higher inside the population already turning to AI tools.

If these figures prompted you to consider support for yourself, the Stress Level Test offers a starting point for reflection and a Stress Diary can organize situations to discuss with a professional. Neither an app's popularity nor a self-test score establishes which care is appropriate for an individual.

The underlying YouGov AI Usage tables identify a sample of 1,082 US adult citizens and an April 1-5, 2026 field period. The question asks whether someone has ever used AI for therapy, so the result does not measure current weekly use or active clinical care.

Who uses AI for therapy: demographics

Age is the clearest dividing line in the data. A study led by RAND researchers with co-authors from Brown University and Harvard Medical School, published in JAMA Network Open on November 7, 2025, surveyed 1,058 US adolescents and young adults ages 12 to 21 in February and March 2025. The sample included children as young as 12, which is why parents weighing AI therapists for kids and teens should look at this data closely. About 1 in 8 reported using AI chatbots for mental-health advice, and usage was higher among the older end of that range, with roughly 1 in 5 respondents ages 18 to 21 reporting use.

Attitude data points the same direction. In the 2026 YouGov survey, adults under 30 were about twice as likely as older adults to say they would feel comfortable working with an AI therapist, at 37 percent versus 20 percent. That comfort gap helps explain why early adoption skews younger, including toward Gen Z.

Beyond age, reliable demographic breakdowns are limited. Most published surveys focus on age and general attitudes rather than detailed cuts by income, geography, or diagnosis, so claims about exactly which groups use AI therapy most should be treated as provisional. Comfort with a hypothetical tool is an attitude measure. It should not be counted as actual use, and the youth sample includes young adults as well as minors.

Why people turn to AI for mental-health support

The reasons people give cluster around access, cost, and comfort. Cost and availability are recurring themes in both the YouGov polling and the Sentio University survey: AI tools are typically cheap or free and available at any hour, which matters when human therapy is expensive or wait times are long. These drivers map onto the broader AI therapy pros and cons: access and comfort on one side, unproven effectiveness on the other.

Reduced fear of judgment is another consistent driver. In YouGov's 2024 polling on AI chatbots and therapy, privacy and anonymity, and the sense that an AI will not judge, came up as reasons people are willing to open up to a chatbot. This lower-stigma framing is part of why some users say they share things with AI that they would hesitate to tell a person.

There are also early signals on perceived helpfulness. In the JAMA Network Open study of adolescents and young adults, among those who used chatbots for mental-health advice, two-thirds engaged at least monthly and 92.7 percent said the advice was somewhat or very helpful. Those percentages apply to respondents who had already used AI for advice, rather than to everyone invited to the survey. The study did not independently verify clinical improvement from the reported advice.

What surveys say about satisfaction and trust

Satisfaction data is genuinely early and mixed, and the evidence base is limited. On the positive side, the Sentio University survey reported that 63 percent of users said large language models had improved their mental health, and the adolescent study found high self-reported helpfulness among users. The same pattern of high self-reported satisfaction shows up in AI therapist reviews.

Public trust, however, remains cautious. In the 2026 YouGov survey, 55 percent of Americans said AI is worse than human professionals at conducting therapy, 14 percent said it is about equal, and 10 percent said it is better. Looking a decade ahead, half still expected AI to be worse than people at therapy, suggesting skepticism is not only about today's tools. Psychotherapy research gives that instinct some backing: a 2018 meta-analysis of 295 studies with more than 30,000 patients found the alliance between client and therapist to be one of the most consistent predictors of whether therapy works, and that relationship is the ingredient a chatbot can only imitate.

Clinicians report widespread but watchful awareness. An American Psychological Association report published in 2026 found that 77 percent of psychologists had spoken with patients who used AI for support or engagement, and 35 percent reported patients using AI as an additional mental-health professional. At the same time, the large majority of psychologists expressed concerns about patients relying on chatbots. The denominator is psychologists answering the survey. These figures do not mean that the same percentages of all patients or all US adults use chatbots.

Concerns and limits in the data

Alongside curiosity there is real worry. YouGov polling from 2024 asked respondents familiar with AI mental-health chatbots about concerns. Reported worries about AI mental-health tools included data safety (46 percent), lack of access to human support (41 percent), the limited emotional range of chatbots (40 percent), and the risk of misdiagnosis (40 percent). These concerns help explain why interest in trying AI does not always translate into sustained use.

The statistics themselves carry limits. Many figures come from single surveys with modest sample sizes, different definitions of 'therapy' versus 'mental-health support,' and self-selected respondents. Self-reported helpfulness is not the same as a measured clinical outcome, and a 92.7 percent helpful rating from users tells you about satisfaction, not about safety or effectiveness. Controlled AI therapy studies measure outcomes very differently from opt-in surveys.

Because this is a fast-moving area, point-in-time percentages can shift quickly as tools and awareness change. These snapshots show reported use and greater interest among younger respondents, with access and comfort recurring in explanations. Establishing growth in use requires repeated surveys with comparable questions and sampling; different studies cannot be combined into a reliable growth rate.

If you are struggling right now

AI tools are not crisis services and they do not diagnose, treat, or cure mental-health conditions. If you are in crisis or thinking about suicide, call or text 988 in the US to reach the Suicide and Crisis Lifeline, available 24 hours a day.

For ongoing support, AI can be a low-cost, always-available supplement, but it is not a replacement for a licensed clinician. If you want human care, you can browse licensed therapists in our directory and compare options that fit your needs.

How many people use ChatGPT for therapy?

The figures above cannot answer that question specifically. YouGov asks about AI use, and the youth study asks about generative AI chatbots for mental-health advice. Their results cannot be relabeled as ChatGPT market share. A defensible ChatGPT-specific estimate would need a survey question naming the product, a defined population, a time period, and a clear distinction between emotional support and general use.

People may use the same chatbot for homework, personal advice, companionship, and coping with distress. Those activities can overlap, so adding separate percentages risks counting a person more than once. A count of messages, downloads, or registered accounts also cannot establish how many unique people sought mental-health support.

How to cite an AI mental-health statistic accurately

Include the source, survey year, population, question, and result in the same sentence. Distinguish when a survey was conducted from when its report was published. YouGov's article combines separate polls; use the underlying data tables when reporting a particular question.

Check the denominator before repeating a percentage. The Sentio result concerns existing LLM users with ongoing mental-health conditions, while APA reports what psychologists heard from patients. The JAMA study concerns adolescents and young adults. Preserving those distinctions makes a statistic useful to readers and harder to misquote.

A study about the relationship between a human therapist and client does not directly compare that relationship with chatbot use. Similarly, self-reported helpfulness cannot show whether advice was accurate or whether harm occurred. Use research on outcomes and safety for those questions, and keep adoption statistics focused on who reported using a tool.