Objective:

To describe when US adults preferred rapid AI responses over slower clinician-reviewed responses and how they viewed mental health support options with different levels of human involvement.

Approach:

• Survey design: Benda et al conducted a 2024 cross-sectional survey of 1,185 US adults recruited using quota sampling.
• Portal messages: For general, logistical, and medical messages, respondents chose between an AI response within 5 minutes and a health-professional-reviewed response within 48 hours.
• Mental health options: Respondents rated likely use of a therapist available in 3 months, a therapist-monitored chatbot available in 1 month, and an immediately available stand-alone chatbot.

Key Findings:

• Immediate AI responses were preferred for logistical messages by 57.4% of respondents and for general messages by 50.6%; 63.1% preferred professional review for medical messages.
• Likely use was reported by 63.9% for a therapist available in 3 months, 50.2% for a therapist-monitored chatbot available in 1 month, and 44.3% for an immediately available stand-alone chatbot.
• Preferences varied among individuals within each message category; more than 40% preferred waiting for review for logistical messages, while more than one-third chose immediate AI for medical messages.
• The survey bundled response time, authorship, and review, so it cannot establish which attribute drove respondents’ choices.

Interpretation:

The findings describe conditional and varied preferences: respondents sometimes favored speed and sometimes human involvement. The mental health results indicate reported willingness to use both therapist and chatbot options, not that chatbots are equivalent to psychotherapy or that unsupervised general-purpose chatbots are safe.

Limitations:

• Response time, authorship, and human review were not independently varied.
• The nature and extent of clinician review were unspecified.
• Mental health choices were hypothetical and may not predict actual uptake.
• The supplied context ends during discussion of subgroup results, limiting what can be summarized about those findings.

Conclusion:

The article distinguishes minimum oversight needed for safety from additional review that patients may prefer. It describes possible oversight approaches, including pretransmission review, review triggered by uncertainty or patient request, retrospective auditing, and rapid escalation; it notes that review is protective when clinicians have adequate context and time and can modify or intervene.