Young people interviewed about a post-meal support group named the mechanism themselves, without being prompted. It stopped you thinking about food. Not processing, not understanding. Occupying attention.

That matches the wider picture. Across 104 studies and 19,734 people, rumination showed the strongest link to eating disorder symptoms of any emotion regulation strategy, at 0.51, against 0.17 for lack of cognitive reappraisal. Repetitive negative thinking in anorexia runs at d = 1.35, the largest of any eating disorder subgroup.

Monaghan & Doyle 2023 · Leppanen et al. 2022 · Palmieri et al. 2021

A study of 24 women with eating disorder behaviours identified eight distinct harms from diet and fitness app design. Green under-budget indicators produced a feeling of achievement, red produced shame scaled to the overage, and gamified elements drove self-competition to eat progressively less each day. The finding that matters most is that all of it held even when the user's stated goal was recovery.

The harm isn't a number sitting still on a screen. It's a loop. Open, log, see the colour, aim lower tomorrow. You don't need to get inside an app to interrupt a loop. You need to get in front of the door.

Eikey 2021 · Simpson & Mazzeo 2017 · Moody et al. 2025

In a randomised experiment with 168 adults, trajectories of perceived control over an unpleasant stimulus predicted emotional outcomes over and above the control people were objectively given. The group who felt they had none showed more helplessness and more depression. It's the feeling of a way out that does the work.

Trauma-informed practice names the same thing. Voice and choice. This population has already had a lot of autonomy removed by treatment, and a tool that adds one more locked door is repeating the injury rather than helping with it. A wall becomes a control fight, and a control fight gets worked around.

Meier et al. 2025 · SAMHSA 2014

Intolerance of uncertainty is the largest measured difference between people with eating disorders and everyone else in this entire literature. Women with eating disorders scored higher at SMD 1.90, and in anorexia specifically 2.16. Put that beside perfectionism at 1.00 and set-shifting at 0.44 and the size of it is clear. Nothing on the market is built on it.

Tapping something without knowing what's behind it is a small deliberate act of tolerating not knowing. It's the smallest possible repetition of the thing that's hardest, and it happens before you've done anything at all.

Brown et al. 2017 · Dahlenburg et al. 2019 · Wu et al. 2014

A fixed expanding circle at six breaths a minute produced the same immediate physiological result as adaptive heart rate variability biofeedback, with none of the hardware. So the plain circle is not the budget option, it's the evidence-backed one.

Heart rate variability is also raised in acute anorexia as an effect of starvation rather than lowered, so a coherence score would reward the illness and hand someone one more number to chase. There is no sensor here and there never will be. A meta-analysis of breathwork found no dose-response relationship, which is the permission to make this one breath long instead of ten minutes.

Tabor et al. 2022 · Jenkins et al. 2021 · Lehrer & Gevirtz 2014 · Fincham et al. 2023

Difficulty identifying and describing emotions is elevated across every eating disorder diagnosis at medium to large effect sizes. A tool that opens with a feelings question locks a large share of its users out at the front door. So every interaction here is a tap on something, never a description of an internal state.

Westwood et al. 2017

Adaptive perfectionism, meaning striving and goal pursuit, is elevated in anorexia against healthy controls at g = 1.24, higher than the maladaptive kind. Points, streaks and completion counts are engineered precisely to exploit that trait. In this population they aren't motivation, they're fuel. Twenty adolescents with anorexia testing a prototype flagged exactly one feature as a problem, and it was streaks.

A review of fifty studies found gamification reliably increases app usage and is not a significant predictor of therapeutic benefit. Engagement without outcomes, bought with a mechanic that harms this group specifically, is a bad trade.

So nothing accumulates. No account, no history, no local storage, no server. Coming back tomorrow starts blank, which means there is nothing here you can fail at.

Dahlenburg et al. 2019 · Sanchez Roman et al. 2025 · Ngabo-Woods et al. 2026

Adults with anorexia underperform across neuropsychological testing at a moderate effect, and body mass index moderates it, so anyone early in refeeding sits at the impaired end. Set-shifting is impaired across eating disorders and worst in the restricting subtype. Anxiety itself consumes attentional capacity and pushes people toward stimulus-driven attention rather than goal-directed attention.

So there is one decision per screen, no wizards, no timed choices, and controls that sit in the same place every single time. Intolerance of uncertainty also means ambiguity is an active cost here, so the page says what it is before you touch anything.

Stedal et al. 2021 · Wu et al. 2014 · Eysenck et al. 2007

Saturation drives arousal far harder than hue does, at a very large effect, and it moves skin conductance rather than only self-report. Desaturation damps affective intensity in both directions, which is the goal for someone who arrives here agitated. Text and controls still hold full contrast, because low saturation is not the same thing as low contrast.

Dim leads because of when this gets opened. Tension climbs across the waking day, around half of the behaviours in one sample fell between six at night and midnight, and sleep is impaired specifically in anorexia rather than across eating disorders generally.

Sensory sensitivity is also raised and survives adjustment for body mass index, while sensation seeking is lower. In adolescent programmes a large share of patients are autistic or have attention deficit hyperactivity disorder. So everything is quiet by default and every bit of motion can be turned off.

Wilms & Oberfeld 2018 · Lin et al. 2024 · Lavender et al. 2016 · Degasperi et al. 2024 · Saure et al. 2022 · Kinnaird et al. 2020

Three separate safe messaging guidelines say the same thing. No weights, no body mass index, no calories, no measurements, no counts of behaviour, no before and after. Scales, tape measures and empty plates are named as prohibited imagery, which rules out progress rings and gauges too, since a ring is a measuring instrument wearing a friendly face. Appearance words including healthy are on the avoid list, which is why nothing here uses wellness language.

Mindframe, NEDC & Butterfly 2021 · Beat · NEDA

There is no randomised evidence that any app improves outcomes in diagnosed anorexia. None. Eikey established that diet app design harms this group, but nothing has established that taking it away helps, and that gap is real.

No trial has shown that tolerating uncertainty in ordinary situations transfers to eating disorder symptoms, which is the assumption the sealed card rests on. Friction before opening an app is an established idea in general digital wellbeing, and none of that work was done in this population.

So treat this as a well-grounded prototype and not a proven intervention. Measure it, and measure harm first. In the closest published analogue to a calming tool in this group, 57% found it relaxing and 43% reported increased distress. Assume a substantial minority gets worse and build the reporting for that before anything else.

Cruz et al. 2025 · Righetti et al. 2025 · Kruger et al. 2025

Not treatment, not a medical device and not a substitute for a person. It's a doorway. Recovery from an eating disorder happens, including for people who have been ill a long time, and it happens with a team rather than with a web page.