AI Companions for the Gaps Between Therapy Sessions

Therapy is structurally discontinuous. You have a session, and then a week passes, and then you have another session. In the week between, your life continues to produce experiences, insights, setbacks, and material. If you're doing the work, you're thinking about what came up in the session, noticing related patterns, sitting with things that were difficult to acknowledge in the room. But you're also just living, which generates its own volume of content: a difficult interaction at work, an anxious night, a moment that felt inexplicably sad, a conversation that went worse than expected. By the time you arrive at the next session, you have a week's worth of material and fifty minutes to work through it. Therapists know this constraint intimately. A common experience is that clients arrive having accumulated so much from the week that the first thirty minutes are spent reporting and the last twenty are the actual therapy. The discontinuity isn't a flaw in the design exactly — weekly sessions are a practical structure that balances depth with frequency — but it does mean that a lot of the productive therapeutic work happens in the gaps, unsupported.

What Happens in the Gaps

In the best case, clients are doing work between sessions: journaling, using skills learned in therapy, noticing patterns, sitting with the questions their therapist has asked. Some clients do this naturally. Many don't, not because they're lazy but because the unstructured gap is hard to navigate without support. Without something to engage with, the material from a session often dissipates. An insight that felt important in the room fades over the week as ordinary life reclaims attention. A habit you were going to try gets put off because there's no one following up. An anxious loop you were going to apply a cognitive technique to runs unchecked for three days before the next appointment. Research from King's College London found that between-session engagement — specifically, the degree to which clients actively engaged with therapy content outside of sessions — was one of the strongest predictors of treatment outcome across a range of therapeutic modalities. The therapy happened in the sessions, but the outcomes were shaped largely by what happened between them.

How AI Companions Work in the Gap

AI companions used as between-session support don't replace the therapist's role. They provide a structure for the ongoing processing that therapy is trying to develop. You can continue working with the material from your last session — bringing it into conversation, noticing how it applies during the week, asking the kinds of questions your therapist would ask if they were there. They also help with the material that accumulates during the week. Rather than letting that material pile up until the next appointment, you have somewhere to take it as it arises. You arrive at your next session lighter — having already processed the surface layer — and more able to do the deeper work that sessions are designed for. This changes the session dynamic in useful ways. The therapist gets a client who has already done some of the work, who arrives with more insight and less accumulated urgency. The session can start further in.

The Continuity Question

One objection to AI companions in a therapeutic context is that they might provide support in ways that undermine the therapeutic work — offering easy comfort where the therapy is trying to build tolerance for discomfort, or validating thoughts that the therapy is working to restructure. This is a real consideration. How you use an AI companion between sessions matters. The useful mode is exploratory — asking yourself the questions your therapist would ask, noticing patterns, engaging with difficulty rather than avoiding it. The less useful mode is seeking reassurance and validation for the same loops that therapy is trying to interrupt. A study from the University of Sydney examining supplementary digital support between therapy sessions found that structured between-session tools produced significantly better outcomes than unstructured ones. The structure mattered. AI companions that help you explore rather than reassure serve the therapeutic work better.

Practical Patterns That Work

Some patterns that work well: reviewing the main themes from a session with the AI in the day or two after, while the material is fresh. Using the AI mid-week when something arises that clearly connects to what you've been working on. Preparing for the next session by articulating what you want to bring and what you noticed since the last one. What works less well: using the AI to avoid sitting with discomfort that the therapy is explicitly trying to help you tolerate, or to get reassurance about things your therapist is trying to help you develop more nuanced responses to. The AI should be in service of the therapeutic direction, not in tension with it. Most therapists, if asked, would have clear views on how a between-session support tool could best serve their work with a particular client — and that conversation is worth having.