Building wellbeing apps responsibly: what we learned from GameHarbor and MindHarbor
Mental wellbeing apps can help people, and they can also overpromise, over-collect and lose users within days. The design decisions we made for our own apps, and the research that shaped them.
Key takeaways
- Research warns that many digital health products fall below quality thresholds and that most users abandon wellbeing apps soon after downloading (Frontiers in Psychiatry, 2025).
- We designed GameHarbor and MindHarbor around short sessions, offline use, on-device data and clear safety routes, and we state plainly that they are not medical services.
- The test of a wellbeing app is whether people feel better, not how long they stay in it.
Why this is hard
A 2025 review in Frontiers in Psychiatry by Babu and Joseph is a useful reality check. Citing other studies, it notes that of the more than 350,000 digital health products available, a large majority fall below established quality thresholds; that most people stop using mental health apps soon after downloading; and that streaks, notifications and gamification can encourage habit rather than real benefit. It also highlights privacy failures across the sector.
The authors recommend clear disclaimers that apps do not replace professional care, built-in routes to professional help, careful consent around data, and independent evaluation. We agree, and those recommendations shaped how we built our own apps.
Decisions we made for GameHarbor
GameHarbor is a calm arcade of 62 short games with an AI companion called Ollie. The design decisions that mattered most:
- Short, put-down-able sessions. Most games take a few minutes and can be left at any moment without losing progress. There is no penalty for closing the app.
- Offline and on-device first. Every game, exercise and an on-device version of Ollie work without a connection, and speech is processed on the phone. Accounts are optional.
- Safety before engagement. When someone appears to be struggling, Ollie does not suggest games. Safety responses point to country-specific helplines.
- Standard tools, honestly framed. Self-check questionnaires such as PHQ-9 and GAD-7 are offered conversationally as a check-in, not a diagnosis.
- Plain disclaimers. The app states that it is a wellbeing app, not a medical or therapy service.
Decisions we made for MindHarbor
MindHarbor takes a different route to calm: an open-world 3D island to wander at your own pace, with relaxing mini-games and spaces for breathing, reflection and winding down.
- Companionship is optional. Players can explore alone, or with a companion who uses CBT-informed techniques such as reframing and small, agreed plans.
- No pressure loops. Collectables and exploration reward curiosity rather than daily streaks.
- Clear boundaries. The game describes itself as a relaxing game and does not offer medical advice, diagnosis or treatment.
A checklist for anyone building a wellbeing app
- Define the benefit in plain terms and do not claim more than you can show.
- Design for short, voluntary sessions. Measure whether people feel better, not only time in app.
- Keep data on the device wherever possible, and make accounts optional or clearly explained.
- Build a safety route: detect distress, stop promotional prompts, and signpost local professional help.
- Use validated questionnaires only as check-ins, with clear wording that they are not a diagnosis.
- State plainly that the app is not a medical or therapy service.
- Review content with qualified professionals before launch and after major changes.
Measuring benefit without dark patterns
Engagement metrics are easy to inflate and easy to misread. For a wellbeing app, better signals include:
- Whether people complete a calming activity, not how long they stay in the app.
- Self-reported mood before and after an exercise, collected optionally and kept private.
- Whether people return by choice, without streak penalties or guilt-based notifications.
- How often safety routes are triggered and whether people follow them to help.
Privacy decisions in practice
Every piece of data a wellbeing app collects is sensitive by nature. In GameHarbor, the journal, moods, check-in results and chats are stored on the device by default, and syncing to an account is optional. Voice is turned into text on the phone, so recordings are not uploaded. The online AI service is optional: the app and an on-device version of Ollie work fully offline without it. Each of these choices removed a category of risk before it could become a problem.
Working with us
We build apps for clients with the same care we apply to our own. If you are planning a wellbeing, health or education app, talk to our app development team.
Questions and answers
Are GameHarbor and MindHarbor medical apps?
No. Both are wellbeing apps designed to help people relax and feel better. They are not medical or therapy services and do not provide diagnosis or treatment.
Why do people stop using mental health apps?
Research cited in Frontiers in Psychiatry (2025) points to app fatigue, unmet expectations and usability problems. Many users abandon such apps soon after downloading.
How should a wellbeing app handle a user in crisis?
It should recognise signs of distress, stop promotional or game prompts, and direct the person to local emergency and mental health helplines.
Can wellbeing apps work without collecting personal data?
Many features can. GameHarbor keeps games, exercises and its on-device companion working offline with data stored on the device, and accounts are optional.
