More Americans are receiving mental health care through their phones or computers than ever before. Video therapy, telepsychiatry, and medication management through audio-only calls have moved from pandemic-era emergency measures to the primary care modality for millions of people dealing with depression, anxiety, ADHD, and substance use disorders.
That expansion has been accompanied by an equally rapid adoption of artificial intelligence tools — specifically, AI-powered clinical scribes that listen to or transcribe virtual appointments and automatically generate the clinical notes that document what occurred.
A 2026 Medscape review of AI-powered clinical note-taking tools used in telehealth settings found significant quality gaps between AI-generated documentation and notes written by human clinicians, raising questions about documentation accuracy, care continuity, and what happens when a summary of a therapy session contains errors no one reviews.
Why This Matters
Clinical notes are not administrative paperwork. They are the backbone of continuity of care. When a patient sees a different therapist, when a prescriber needs to understand what a patient said three sessions ago, when a patient appeals an insurance decision, or when a legal matter requires medical records, the accuracy and completeness of clinical documentation is critical.
Mental health notes are particularly sensitive. They often capture disclosures about trauma, suicidal ideation, relationship dynamics, and medication side effects. If an AI scribe misrenders, misses, or incorrectly summarizes these details, the clinical and legal consequences can be serious.
Yet there are currently no regulatory standards governing the use of AI scribes in telehealth. The FDA has not issued guidance specifically addressing them. State licensing boards have not uniformly addressed documentation requirements in AI-assisted virtual care.
What We Know So Far
Mental health has emerged as the single largest driver of telehealth growth in 2026, according to industry tracking data. The growth has been dramatic: virtual psychiatry platforms, digital therapy apps, and general telehealth services with mental health offerings have expanded coverage to millions of Americans who previously had limited access to in-person care.
At the same time, virtually every major telehealth platform has integrated or is actively integrating AI-powered ambient documentation — systems that listen to the appointment in real time and generate a structured clinical note automatically.
The Medscape review assessed note quality across multiple dimensions, including accuracy of symptom description, capture of patient-expressed concerns, documentation of medication changes, and appropriate recording of safety assessments. It found that AI-generated notes showed meaningful deficiencies in nuanced content, particularly in areas where the clinical significance of what was said depends on context, tone, or patient disclosure that was not explicitly stated.
Where the Risk Is Greatest
The documentation quality concern is most acute in:
- Mental health encounters involving safety assessments — conversations about suicidal ideation, self-harm, or harm to others, where precise clinical documentation has legal and ethical significance
- Medication management visits, where exact dosage discussions, side effect descriptions, and prescribing rationale need accurate records
- Patients with complex diagnostic pictures, where notes that miss nuance can mislead future providers
- Telehealth platforms with high patient volume and limited human review of AI-generated notes
The risk is lower in straightforward follow-up appointments for stable conditions where the note is primarily administrative.
What Doctors and Experts Say
Clinicians who have tested AI scribes note that the tools vary significantly in quality — and that even the best AI scribes require human review to catch errors. The problem is that when clinical notes are auto-populated immediately after a session, the provider may not take time to review them carefully, particularly in a high-volume telehealth practice.
Psychiatrists and therapists have also noted that AI scribes may handle certain content differently than expected — glossing over a patient's mention of passive suicidal ideation, for example, because the phrasing was indirect. In a clinical context, that is an unacceptable documentation gap.
Regulatory bodies have acknowledged the issue without yet resolving it. CMS and the FDA are both watching the space, but specific standards for AI scribes in behavioral health telehealth do not yet exist.
What the Evidence Shows — and What It Does Not
The Medscape review is an observational quality assessment — it documents a pattern of performance gaps but is not a controlled clinical trial. The findings reflect the current state of AI documentation tools available in the telehealth market, not what AI documentation might eventually achieve with further development.
The review does not quantify the direct patient harm attributable to AI scribe documentation errors — that causal link is difficult to establish. What it establishes is that the documentation produced by AI scribes diverges meaningfully from what human clinicians write in important clinical domains.
Who Is Most Affected?
- Patients receiving mental health care through telehealth platforms that use AI scribes without transparent disclosure
- Patients with complex mental health histories for whom accurate documentation of their full clinical picture is especially important
- Patients managing psychiatric medications, where documentation of dosage, rationale, and side effects affects future prescribing
- Patients whose care is delivered across multiple providers or platforms, making accurate records the only thread of continuity
What You Can Do Now
- Ask your telehealth provider whether AI-powered note-taking is used in your appointments, and whether a human clinician reviews the notes before they are finalized in your medical record.
- Request access to your clinical notes through your patient portal. Most health systems and telehealth platforms are required to make notes available to patients. Reading them can help you catch errors.
- If you notice inaccuracies in your notes — particularly related to medications, symptoms you reported, or safety conversations — contact your provider to request a correction.
- If you have concerns about how your mental health information is being documented, ask your provider to describe their documentation process and any quality review steps in place.
- When choosing a telehealth mental health provider, asking about their documentation practices and whether notes are clinician-reviewed is a reasonable and appropriate question.
Cost and Access: What Patients Should Know
Telehealth mental health services have dramatically expanded access to care for people without local providers or with transportation barriers. That access benefit is real and should not be dismissed because of documentation concerns.
The quality gap is a signal that oversight needs to improve, not that patients should avoid telehealth. Patients who are receiving high-quality care through telehealth should continue doing so while advocating for better documentation standards.
What Happens Next
Regulatory guidance on AI in clinical documentation is expected from CMS and from the FDA in 2026 or 2027, though no firm timeline has been announced. Professional organizations, including the American Psychiatric Association and the American Psychological Association, are developing position statements on AI-assisted documentation. MedicalDaily will track regulatory developments in this space.
The Bottom Line
Telehealth has transformed access to mental health care — and that is a genuine public health benefit. But the AI scribes now used to document those virtual visits have documented quality gaps that no one has yet established regulatory standards to address. Patients receiving mental health care through telehealth should know that this gap exists, request access to their notes, and advocate for human review of AI-generated documentation.
