Most mental wellness apps make the problem worse.
The category is loud, and it is failing. Median 30-day retention for consumer wellness apps sits under 5%, and the apps that hold attention do so by methods that look, on inspection, like the architecture of other dependencies: streaks that punish absence, scores that quantify a bad week into a number a person can fail at, push notifications timed to interrupt rather than serve. The interface inherits the grammar of the slot machine, then dresses it in calm pastels.
Behind that interface, the content is rarely matched to the person. Most users meet the same library of meditations, gratitude journals, and breathing exercises that every other user meets. The app does not know whether the week behind you was the worst in a year or the easiest. It serves the same prompt either way. When the prompt does not land, the user concludes that they have failed the practice. They delete the app, and the failure becomes private.
The cost of that failure is large and increasingly visible. Mental ill-health is now estimated to cost UK employers £56 billion a year in lost productivity, presenteeism, and turnover; the social cost is much larger. Among adults aged 25 to 45, the cohort most likely to download a wellness app, anxiety and depression continue to rise. The category has had a decade to help, and the trend has not bent.
Morrígan exists because the people most likely to be helped by a well-designed companion have been served, instead, by the architecture that broke their attention in the first place. We wanted to build the opposite of that.
The lifestyle medicine framework.
The clinical ground under Morrígan is lifestyle medicine, the evidence-based practice that sits adjacent to, and increasingly inside of, primary care. The American College of Lifestyle Medicine codifies the field around six pillars: sleep, stress, nutrition, physical activity, connection, and the substances and behaviors people consume to regulate themselves. Each pillar has a substantial literature linking it to mental health outcomes, and each is something a person can act on without a prescription.
The framework is not a self-help heuristic. It is the structure used by clinicians to address whole-person health, and the evidence base behind it is substantial: the 2024 BMJ meta-analysis of 218 trials on exercise and depression, the Harvard Study of Adult Development on connection, the rapidly accumulating literature on the gut-brain axis and the metabolic load of poor sleep. None of this is novel. The novelty is the decision to let a daily companion sit inside it.
Each pillar is detailed on the home page. Briefly: sleep is the ground of everything else. Stress is acute or chronic, and the difference matters. Nutrition affects mood and cognition through mechanisms now well characterized. Physical activity is one of the most reliable mental health interventions available. Connection is a physiological requirement. Substance is the pillar that names, without moralizing, the substances and behaviors a person turns to when regulation is hard.
We did not invent a framework. We built a companion that lives inside a well-tested one.
The AI at the center.
The companion is held together by an editorial layer we call the Explainer. It works at two scales. The daily scale is short-form: a sentence or two of observation written in response to what the user has shared in the last 24 hours. The weekly scale is long-form: a per-pillar analysis that reads across the week, identifies the patterns that have settled or shifted, and offers a reading of what the data appears to suggest.
The voice is editorial, not coaching. Coaching presumes a goal the user has agreed to and a method the coach is qualified to apply. The Explainer presumes neither. It observes, in the register of a careful editor, what the week appears to contain. The user decides what to do.
Provenance is built into the surface. When Morrígan writes that the user's sleep is settling more deeply this week, the observation is a link. Tapping it opens the specific data points behind the sentence: the bedtimes, the wake times, the wind-down activities the user has recorded, the absence of late caffeine. Nothing the AI says is unmoored from the data the user has provided. The provenance is not a feature for power users. It is the only honest way to write about another person's life.
Designed to quiet.
Most consumer software is built to maximize the time the user spends inside it. Morrígan is built to do the opposite. We hold a five-stage progression model: a user begins with frequent contact, learns the practices that anchor for them, and over months reaches a state where the companion's presence in their day is light. Six withdrawal triggers, written into the system, instruct the AI to surface less rather than more as specific patterns stabilize. Consistent sleep for eight weeks reduces the cadence of sleep-pillar surfacing. A settled connection signal reduces prompts to reach outward.
This is at odds with how the wellness category measures success. We do not measure daily active use. We measure whether the patterns the user came to address are settling, and whether, over time, the companion is needed less. A user who comes back lightly, every few weeks, because something has shifted and they want a read on it, is the design outcome. A user who opens the app every morning at 7:32 is not.
The economics of this choice are uncomfortable for a venture-backed consumer app. They are not uncomfortable for a clinically grounded product designed for a long life. The contrast is the point.
What this is, what this is not.
Clarity here matters more than rhetoric. Morrígan is:
- A daily companion grounded in lifestyle medicine
- An editorial reading of patterns across the six pillars
- A private surface, designed to surface less over time
Morrígan is not:
- A clinical tool, a diagnosis, or a replacement for one
- A substitute for therapy, medication, or medical care
- A productivity hack, an optimizer, or a quantified-self device
- A crisis service. People in crisis should call their local crisis line
Saying what a product is not is a deliberate act. The categories the product could be mistaken for are categories where the wrong reading does real harm. Morrígan is a companion, in the long sense of that word. The companion has limits, and the limits are part of the design.
Mental health products need long product lives. We are building for one.
Morrígan launches in September 2026. The work between now and then is a careful one. We are building the clinical content with practicing clinicians, the editorial voice with the editors and writers we trust, and the system with engineers who understand that a product which holds a person's mental health data carries an obligation that most consumer software does not. We expect Morrígan to be used, by the people it suits, for years rather than months. The decisions we are making now are made under that assumption.