Why People Are Using AI for Mental Health (And When It’s Not Enough)

More Americans than ever are turning to a chatbot before they call a therapist’s office. That shift is accelerating in ways the numbers make hard to dismiss. 

According to the BetterHelp 2026 State of Stigma Report, a survey of 2,000 U.S. adults fielded in March 2026, 15% of Americans currently use AI for mental health support, a figure that has climbed six percentage points in a single year. 

For a country where 74% of people believe society still discourages seeking mental health care, the pull of a tool that is private, available at any hour, and often free, makes a certain kind of sense.

The question worth asking is not whether AI belongs in the mental health conversation. The data says it already does. 

The more useful question is where that role has real limits, and why a licensed therapist remains central to any mental health journey that goes beyond coping with a rough week.

Why people are turning to AI for mental health care

The three most common reasons Americans turn to AI for mental health support (general anxiety, financial stress, and work-related pressure) share something in common. 

These are concerns that tend to surface outside of business hours, carry a degree of shame, and do not wait politely for a scheduled appointment. A person lying awake at 2 a.m. cycling through money worries is not well-served by a system that requires waiting two weeks for availability.

Cost, convenience, and the reality of right now

Among Americans who have used AI for mental health support, 54% say it is more affordable than other options, and 47% value its capacity for immediate response, per the 2026 State of Stigma Report. 

For people who cite cost as a barrier to therapy, 52% of non-therapy-seeking Americans in the same survey say a low-cost or free tool that responds in seconds carries genuine practical appeal.

The broader access problem matters here. In November 2025, the American Psychological Association issued a health advisory acknowledging that the low cost and accessibility of AI chatbots and wellness apps have made them a go-to resource for people who cannot find or afford care from licensed providers. AI is not creating that gap; it is filling one that already exists.

What AI can offer in the moment

The strongest case for AI in mental health is in-the-moment support: tools that help someone manage acute distress, stay engaged with their mental health between professional appointments, and reduce the friction between recognizing a problem and doing something about it. 

For people navigating everyday stressors, a difficult conversation at work, a period of disrupted sleep, or a wave of anxious thoughts, AI-based tools can offer breathing exercises, journaling prompts, and basic stress management guidance may provide real, if limited, relief.

At least 50% of Americans who have used AI for mental health purposes report a positive impact across areas including grief, stress, and sleep, according to the 2026 State of Stigma Report. 

For everyday challenges, those are meaningful outcomes. AI can also serve as a low-stakes space for someone who is not yet ready to commit to professional care, a place to name what they are feeling before they are willing to bring it to a therapist.

Where AI falls short and why licensed therapists are critical

There are certainly some positive aspects related to AI tool usage in the mental health space that are innovative and exciting, and offer potentially practical solutions to unique situations.

However, in a space as nuanced as mental health care, the need for licensed therapists is critical to long-term success in a way that technology alone cannot replace. Human elements that cannot be replicated 

The same data that documents AI’s growing role in mental health also makes its limits explicit. When Americans who have used AI for mental health support were asked which source of care they prefer for the qualities that matter most, discretion, emotional nuance, comfort level, and communication depth, they consistently chose human therapists over AI. 

Therapy is built on a relationship. The therapeutic alliance, the trust and sustained rapport between a client and a licensed provider, is one of the most consistently cited predictors of positive clinical outcomes in the research literature. 

A chatbot can process language and generate responses that feel supportive. It cannot register the hesitation behind a sentence, connect something said in a current session to a pattern that emerged months earlier, or calibrate its clinical approach to the full, lived complexity of one individual’s history.

The 2026 State of Stigma Report makes the distinction plainly: people prefer AI for affordability and availability. They prefer therapists for discretion, communication depth, and the emotional attunement that drives outcomes.

The clinical limitations are significant

Beyond the question of empathy, there is the matter of clinical authority. 

Licensed therapists are trained to assess mental health concerns, identify patterns that develop across sessions, and use evidence based modalities, including cognitive behavioral therapy, to support individualized care."

They are accountable to professional licensing boards and equipped to recognize when a client’s situation calls for a higher level of care.

AI tools, including the most sophisticated large language models available today, hold no licenses, cannot medically diagnose, and operate outside the regulatory and ethical structures that govern professional mental health practice. 

The APA’s November 2025 health advisory was clear on this point: most generative AI tools and wellness applications currently lack the scientific validation and oversight necessary to ensure user safety in mental health contexts, and even tools built with strong psychological science do not yet have sufficient evidence of efficacy or clinical safety.

For someone managing chronic depression, processing trauma, or working through persistent patterns of disordered thinking, responses that address the surface of a problem are not a substitute for structured clinical care. 

The concern is not only that AI support may fall short; it is that reliance on it may delay a person from pursuing the kind of care that addresses root causes.

Privacy, trust, and the depth of what gets said

One of the less obvious reasons human therapists matter is the quality of self-disclosure they make possible. In the 2026 State of Stigma Report, Americans identified privacy, comfort level, and concern about how others might respond as areas where they still prefer a therapist to AI. 

These are not minor preferences. The depth to which a person will open up about their mental health is inseparable from the trust they have in their provider.

A skilled licensed therapist can create an environment in which a client can say things they have never said to anyone, and can do something clinically meaningful with what is shared. That process is categorically different from exchanging messages with an interface, regardless of how private that interface may feel.

It’s important to note that the AI and mental health field is constantly evolving. Research is growing around positive mental health outcomes when engaging with an AI agent. A 2023 clinical study concluded that participants had a meaningful AI-based conversation with an agent, helping to reduce psychological distress, specifically symptoms of depression and general distress.  

However, the evidence base remains limited, with concerns around safety reporting, methodological rigor, unclear mechanisms of improvement, and insufficient long-term follow-up data.

Why using both AI and therapy may work best

The most productive frame for AI in mental health care may be neither “replacement” nor “supplement” but simply a starting point. 

AI as a starting point, not a destination

For people who have never engaged with therapy, who are unsure whether what they are experiencing warrants professional attention, or who are not yet ready to commit to a structured therapeutic relationship, AI can lower the threshold for engaging with mental health at all.

Millennials illustrate the point well: 42% of Millennial respondents in the 2026 State of Stigma Report describe AI as a gateway or supplement to human care, the highest share of any generation. Getting someone from “something feels wrong but I don’t know what to do about it” to “I’ve made an appointment with a therapist” is a meaningful step, and AI may help bridge that gap.

That framing also reflects what most Americans actually believe: 46% of respondents in the 2026 report agree that AI services for mental health are most effective when paired with care from a medical professional. That is not a rejection of AI — it is a recognition that the two forms of support are suited to different needs and that their combined value exceeds what either delivers on its own.

Building a personalized and accountable path to care

What a licensed therapist provides that AI cannot is structure over time: a consistent relationship, a documented clinical history, a treatment plan tailored to a specific individual’s diagnosis and goals, and the judgment to recalibrate when something is not working. That kind of care compounds. Each session builds on the last.

Working with a therapist through an online therapy platform, like BetterHelp, connects individuals with licensed professionals who bring a minimum of three years and 1,000 hours of clinical experience, and whose work is grounded in accountability structures AI tools simply do not have. 

Among BetterHelp users:

• 72% experienced a reduction in symptoms within their first 12 weeks of treatment
• 69% showed meaningful improvement in anxiety and depression
•  62% reached full remission, outcomes that reflect what structured, relationship-based care delivers at scale.

AI-based tools can complement that work: providing a journaling resource between sessions, offering in-the-moment grounding exercises, or simply reducing the inertia that keeps people from seeking help in the first place. 

Accountability, clinical depth, and a tailored plan, though, belong to the therapeutic relationship.

AI has earned a serious place in the mental health conversation. For people facing real barriers of cost, access, geography, or time, it offers something concrete: low-friction support that is available when traditional care is not. For many people, it may also be the first step toward engaging with mental health at all, and that matters.

What AI cannot provide is the clinical expertise, sustained human connection, and tailored structure that make therapy effective for conditions that run deeper than everyday stress.