Relationship first: what an app can learn from good therapy and where its limits lie

People who start therapy rarely say in the first session what it is really about. Trust has to grow first: the feeling of being heard and understood. Without this relationship, even the best techniques barely take hold. For an app, that is an honest challenge. This article shows what research says, where the limits are, and what we try to do anyway.

How clarifying, interpreting and working through resolve inner conflicts is the topic of our article Resolving inner conflicts. This one is about what comes before.

Why the relationship matters so much

Psychotherapy research speaks of the therapeutic alliance or working alliance. In Edward Bordin’s influential model it has three parts: agreement on goals, agreement on the shared tasks, and a solid bond¹.

How much this matters shows in a large meta-analysis by Flückiger and colleagues covering 295 studies with more than 30,000 patients. A good alliance was robustly linked to better therapy outcomes (r = .28). Remarkably, in internet-based psychotherapy the link was practically the same size (r = .28)¹.

Empathy is similarly well documented. A meta-analysis of 82 samples with more than 6,000 clients found empathy to be a moderately strong predictor of therapy outcome across different schools of therapy². The empathy that clients themselves perceived predicted outcome best². What counts, then, is whether you feel understood.

What an app cannot do

Honesty matters here. MindMotion is not psychotherapy. An app can neither imitate the setting of therapy nor replace the relationship with a person who accompanies you for months, reads your facial expressions and takes responsibility.

Professionals name these limits clearly. In studies of text-based digital therapy, therapists worry that missing nonverbal cues make communication harder. They also feel a special responsibility, especially in crises without physical proximity⁴. In November 2025 the American Psychological Association published a health advisory on AI chatbots and wellness apps. It states that such tools should not be used as a replacement for a qualified professional⁶. The APA also warns of three typical risks: chatbots that mostly agree with their users, unhealthy dependence, and unreliable handling of crises⁶.

We treat these points as guard rails, not as a footnote.

What is possible digitally

Despite these limits, research shows that a good working relationship can form through text, too. A systematic review of text-based digital psychotherapy came to exactly this conclusion. In most studies that examined it, the alliance was linked to treatment outcome⁴. The authors suggest that writing gives time to reflect and reduces the feeling of being observed, which may foster openness⁴.

Even with a purely digital conversational agent, users in a large observational study reported a noticeable bond within a few days⁵. This study has clear limits, however. It was carried out by the app’s own developers, had no comparison group and measured only one point in time⁵. It shows that a bond is possible, not that it equals a human relationship.

Which techniques can be carried over

Listening and reflecting back

Good listening means giving back what you heard so that the person feels understood. Motivational interviewing uses reflections for this. With ambivalence, both sides are deliberately reflected back, what speaks for a change and what speaks against it⁷. That way, nobody feels pushed.

Understanding instead of judging

Empathy can take different forms in therapy². For an app this means staying with your own words, asking rather than interpreting, and naming feelings without judging them.

Agreeing on goals and tasks together

Flückiger and colleagues recommend agreeing on goals and tasks early and regularly assessing the quality of the relationship from the client’s perspective. A strong alliance is often the result of negotiation. It deserves the same attention in internet-based psychotherapy¹.

Addressing ruptures openly

Every relationship has moments when things get stuck. In therapy these are called ruptures in the alliance. The recommendation is to address them directly and promptly¹. Studies suggest that repaired ruptures are associated with better outcomes³.

Our ideas for MindMotion

From these principles we derive concrete ideas. Some are already in the prototype, many are planned and will be tested with users.

• “Did I understand you correctly?” The AI sums up in your own words what it heard, and you correct it with a tap. Not the app is right, you are.
• Negotiating goals, not prescribing them. The app suggests your weekly goal and Moves, and you choose, change or discard them. That reflects agreement on goals and tasks.
• The relationship pulse. After an exercise the app briefly asks: “Do you feel understood?” and “Is the pace right?”. Your answers steer the tone and depth of the next steps.
• Noticing and repairing ruptures. If you skip exercises or express frustration, the app says so openly and offers another way instead of simply carrying on.
• Honest, not flattering. The AI is designed not just to agree but to kindly offer a second perspective, as a counterweight to the risk the APA warns about⁶.
• Continuity with privacy. The app remembers your themes so you don’t start from scratch each time. You decide what is stored. The safest setting is the default.
• Your tone, your pace. Matter-of-fact or warm, short or detailed: you set how the app talks to you.
• A bridge to people. The app makes clear that you are talking to an AI. It shows crisis services right away and helps you share your insights with a therapist or coach.

Our aim is not to replace therapy. We want to meet people where they are, with the care known from good therapy and good coaching. You can find the concept behind it on our home page. Which inner patterns may lie behind your conflicts is the topic of our article Inner conflicts.