A recent survey conducted by ICANotes found that nearly 40% of clinicians reported that patients had used AI tools before accessing professional support. The tidal wave of AI on everyday life, bolstered by these stats, suggests that using AI in this way is slowly becoming the norm.

This behavior could have sizable implications for risk assessment, initiation of treatment, and the integrity of clinical records. Let’s explore some things clinicians should be aware of in this AI-powered world.

Emotional Support in 2026: Why Are People Turning to AI?

Essentially, some patients turn to AI because therapy feels expensive or difficult to access; others turn to AI because they do not yet recognize that their symptoms warrant professional intervention.

For clinicians, this could highlight AI as a potential indicator of access. When nearly 40% of patients report using commercial AI before their first appointment, the question is no longer whether AI is part of people’s emotional support, but rather how its use could affect the care, symptoms, and mindsets, informed by misinformation, that patients could bring with them.

AI: Supplement or Risky Substitute?

There is, arguably, a role for AI as a tool to complement mental health care. For example, purpose‑built chatbots, tested in controlled settings, could produce short‑term reductions in anxiety and depressive symptoms for some. Even general‑purpose tools can offer self‑reflection, ideas for coping, and a sense of being heard when other care is unavailable. For patients on waitlists or between sessions, this can feel therapeutic.

The risk emerges when general AI becomes a complete substitute for professional care, particularly in acute or complex situations.

In the ICANotes survey, 24.14% of providers reported treating patients whose mental health conditions deteriorated following unguided use of AI tools. That figure aligns with growing concerns in the literature about general‑purpose chatbots offering confident but inaccurate advice, reinforcing distorted cognitions, and failing to escalate safely in high‑risk situations.

Emotional Health Risks

Several risk domains are particularly salient for clinicians evaluating the impact of AI use on patient care, such as:

• Misguided validation. Most AI platforms are designed to be agreeable. For vulnerable users, this can reinforce unhelpful beliefs, encourage avoidance, or minimize the seriousness of risky behavior, which is especially risky when it comes to severe states such as mania or psychosis.
• Crisis‑response gaps. General‑purpose AI is not reliable for suicide triage or crisis counseling. Conversations may continue despite concerning content, and escalation to human crisis resources is inconsistent.
• Dependency and social withdrawal. Heavy daily use could be linked to greater loneliness, dependence, and less socializing, particularly among users who are already socially isolated.
• Privacy and confidentiality. Many direct‑to‑consumer AI tools operate outside HIPAA. This means that sensitive disclosures entered into a chat may be stored, used to train models, or exposed in a breach.

New Questions in Light of AI Use?

For clinicians, AI use could be treated as a routine element of creating a clinical picture as this landscape evolves. For instance, it could be put alongside questions about substance use, social media, or other digital behaviors that affect mindsets.

These might look something like the following:

• “Are you currently using any AI chatbots or apps for emotional support or mental health advice?”
• “How often do you use them, and what do you find most helpful or concerning about those conversations?”

Finding out frequency, specific tools, and any red‑flag interactions allows you to contextualize symptoms, assess dependency patterns, and document potential risk factors in the social history or risk assessment sections of the Electronic Health Records.

How Lower Capacity Means More AI

Other findings from the same ICANotes survey include:

• More than 40% of behavioral health providers spend 11 or more hours each week on documentation, billing, and other non‑clinical administrative tasks.
• More than a quarter of clinicians have reduced their caseloads over the past year to accommodate administrative work, while nearly half say they could see more patients if documentation requirements were reduced.
• Nearly half of clinicians have either dropped at least one insurance plan or are actively considering doing so, and more than 70% say current reimbursement structures do not adequately support quality behavioral healthcare.

When administrative hurdles limit capacity and reimbursement doesn’t align with quality care, patients can face longer waits, reduced access, and higher out‑of‑pocket costs. In that environment, commercial AI becomes a tempting option for emotional support.

In this context, it is important for those in behavioural health to be aware of how this new and increasingly popular form of emotional support could affect their practice.

Survey statistics cited above are from an ICANotes provider survey on patient use of commercial AI tools for emotional support and on administrative burden, reimbursement, and insurance participation in behavioral health practice.

ICANotes is the premier clinical specialty EHR for behavioral health. Designed by a psychiatrist, it features unique clinical content enabling clinicians to create comprehensive, compliant charts faster than any other system. Learn more at ICANotes.com.