Purpose-built vs. general-purpose AI: How Wayhaven builds better digital wellness support

In an era in which wellness support is increasingly difficult to access, many people are turning to AI chatbots for emotional support and guidance. A new study from Stanford University researchers (Scholich et al., 2024) offers important insights into how these systems actually perform.

The researchers conducted an analysis of how different general-purpose AI chatbots respond to wellness scenarios compared to licensed therapists’ approaches.

Where general-purpose AI shows promise 

The study found that chatbots demonstrated:

• Strong validation and normalization of user experiences
• Consistent empathetic responses
• Effective delivery of psychoeducational information

Key areas for improvement

1. The advice trap: The study found that general-purpose chatbots defaulted to giving direct advice rather than using Socratic questioning to help users develop their own insights. This bypasses a crucial self-discovery process that contributes to lasting change.
2. Resource connection gaps: Most chatbots failed to effectively connect users with appropriate support resources, with a minority providing specific resource information.
3. Engagement barriers: While AI companions provided emotional support, they often failed to foster the active skill-building and practice that drive meaningful improvements in well-being.
4. Cognitive restructuring challenges: The chatbots showed limited ability to guide users through effective thought reframing exercises.
5. Memory constraints: Current general-purpose chatbots have limited ability to retain and use information about previous conversations, restricting their ability to provide personalized support.

How Wayhaven addresses these challenges 

Our purpose-built AI mental wellness coaching platform specifically addresses these limitations through:

• User agency and self-efficacy through concrete skill practice - research consistently shows that self-generated insights lead to more lasting change than externally provided solutions
• Socratic questioning that guides users toward their own insights rather than defaulting to ineffective advice-giving
• Robust resource connection protocols that not only connect users with appropriate support but actively promote utilization
• Active skill-building exercises that drive behavioral change through guided practice and mastery
• Structured cognitive reframing support with appropriate scaffolding, contributing to decreased endorsement of automatic thoughts and intensity of negative emotions
• Conversational memory that tracks progress over time and enables truly personalized interactions
• Clear boundaries that foster independence while maintaining appropriate support

The future of AI in user-centered wellness support lies not in general-purpose chatbots, but in purposefully designed systems that combine AI’s consistency with evidence-based mechanisms to cognitive and behavior change. At Wayhaven, we’re committed to leading this evolution through intentional design that puts user empowerment and evidence-based practices at the center of everything that we do.

‍

‍

References

‍

Scholich, T., Barr, M., Stirman, S. W., & Raj, S. (2024). Can chatbots offer what therapists do? A mixed methods comparison between responses from therapists and LLM-based chatbots [Unpublished manuscript]. Institute for Human-Centered AI, Stanford University.

‍