Designing for Anxious Users: UX Lessons From Digital Health

Most apps open with a value proposition and a sign-up prompt. For an anxious user, being asked to commit before understanding what will happen is a reason to leave.

What worked better was leading with what the product will and will not do. Plain language, no clinical vocabulary, no marketing tone. Something closer to "this will ask a few questions about how you feel about dental visits, and then explain what usually helps" — before any account request.

The measurable effect is on completion rather than acquisition. People who understand the shape of what is coming are far more likely to finish it.

Standard form design says group related fields to reduce perceived length. For an assessment that touches something someone is anxious about, a dense screen of questions reads as an interrogation.

One question per screen tested better despite requiring more taps, for reasons that are emotional rather than functional:

• Each answer feels complete rather than partial.
• There is a natural pause between questions.
• The user is never confronted with the full scope of what is being asked.

The trade is real — it is slower, and for a non-anxious user it would be worse. Knowing which user you are designing for is the whole point.

On progress indicators

We expected to remove them entirely. What worked was showing progress without a total: "question 4" rather than "4 of 12". The user knows they are moving; they are not confronted with how far is left. Removing the indicator entirely made people uncertain whether the process was endless, which was worse.

Tone in a health-adjacent product is not a copywriting preference. We wrote it as specification and reviewed against it:

• No diagnostic language. The product describes and supports; it does not assess or label. This is a legal and ethical boundary as much as a tonal one.
• No minimising. "Don't worry" and "it's easy" tell an anxious person their feeling is unreasonable. Acknowledge, then help.
• No urgency devices. Countdowns, streak pressure, and "you're falling behind" messaging are actively harmful here, however well they perform elsewhere.
• Second person, present tense, short sentences. Cognitive load is already high.
• Every result includes what to do next. Information without a next step increases anxiety rather than reducing it.

Having these as written rules meant tone survived handoffs between designers, engineers, and the client, rather than drifting with whoever wrote the last string.

Anxiety is closely tied to a feeling of not being in control. Several small affordances did disproportionate work:

• Always allow going back and changing an answer, visibly. People answer more honestly when they know it is not final.
• Allow skipping with a neutral option rather than forcing a response. A forced answer is worse data than a skip.
• Save and resume by default. Leaving mid-assessment should be a pause, not a loss.
• No auto-advance. Selecting an option should not instantly move the screen — it removes the moment to reconsider and feels like being rushed.

Each of these makes the flow slower on paper. Together they make it completable.

For scoring products the temptation is to lead with the score. For an anxious user, a bare number invites the worst interpretation.

What worked was structuring every result the same way: what this means, what it does not mean, and what you can do next. The "does not mean" section was the one users found most reassuring, because anxiety fills gaps with worst cases and explicitly closing them helps.

We also avoided colour-coding results red. It is an efficient visual convention and it reads as alarm. Neutral tones with clear text carried the same information without the spike.

These lessons apply well beyond health. Any product used in a stressed state — financial hardship tools, legal help, incident reporting, retirement planning like Rentenkompass — benefits from the same shifts:

1. Explain what will happen before asking for commitment.
2. Pace input rather than batching it.
3. Write tone rules as requirements, not preferences.
4. Give control: back, skip, pause, resume.
5. Pair every result with a next step and an explicit boundary on what it does not mean.

The underlying principle is simple. Efficiency is a proxy for a good experience, not the definition of one. When you know the user's emotional state, design for that instead.