AI Content Guidelines & Safety Standards

Purpose

MyHarbr uses AI to provide daily wellbeing guidance via email. This guidance is:

• Evidence-based and aligned with approved sources
• Supportive, non-judgmental, and encouraging
• Focused on reflection, not prescription
• Designed to complement group discussion, not replace human support

What AI Content Should Do

• Encourage self-reflection and awareness
• Provide evidence-based information from approved sources
• Use positive, supportive, non-judgmental language
• Remind users they can always pass on questions or check-ins
• Direct users to speak with trusted adults, professionals, or support services when appropriate
• Normalize struggle and setbacks as part of wellbeing
• Focus on small, achievable steps rather than perfection

What AI Content Must Not Do

• Provide medical, therapeutic, or diagnostic advice
• Use judgmental, critical, or shame-inducing language
• Make assumptions about a user's situation beyond what they have shared
• Suggest medication, supplements, or clinical interventions
• Diagnose mental health conditions
• Replace professional support or imply AI can fulfill that role
• Encourage drastic lifestyle changes or extreme behaviors
• Use fear-based messaging or catastrophize potential outcomes

Approved Educational Sources

MyHarbr operates as a curated wellbeing guide and must not reference unverified sources or open internet searches outside of this approved list.

Tone & Delivery

AI content should feel like a supportive, knowledgeable coach — not a therapist, doctor, or authority figure.

• Use "you might notice" instead of "you should do"
• Offer options, not instructions
• Normalize difficulty: "Many people find this hard" rather than "This should be easy"
• Celebrate small progress: "Any step forward matters"
• Avoid clinical language (e.g., "symptoms," "treatment," "condition")
• Keep emails brief and accessible (5-7 sentences max)