* Image created with AI by kikasworld.com
Gestalt therapy was built on a deceptively simple idea: real contact happens only in the present moment, between two people willing to actually meet each other. Martin Buber called this an I-Thou relation — meeting the other as a whole, living presence, not as a case, a category, or a predicted outcome. The moment you relate to someone through what you already know about them instead of who they are right now, you slip into what Buber called I-It — relating to an object of use, analysis, or memory, rather than a presence that can address you back.
Which raises an uncomfortable question for anyone who has ever sat across from a client with a folder of notes on their lap: if the therapist walks in already “knowing” the client, has the real encounter already been compromised before it begins?
And more recently, a stranger question has joined it: what happens when the one taking the notes isn’t a person at all?
When Memory Becomes the Enemy of Meeting
Let’s hold up a few different lenses to the human side of this first.
The purist relational view. Thinkers like Yontef and Hycner would argue that each session is, in principle, a new organism-environment field. The client today is not identical to the client from last week — their figure/ground has shifted, their self-regulation has moved on. A therapist who enters the room already anchored to last session’s narrative risks imposing a fixed “who you are to me” onto someone who may have already grown past it. In this view, memory can become a subtle form of violence — a refusal to let the client be new.
The “ground, not figure” view. A softer, more workable position: memory doesn’t need to disappear, it just needs to move to the background. Buber never demanded amnesia for genuine dialogue — what he demanded was a quality of attention, an openness to being surprised. A therapist can hold prior sessions the way any good listener holds context: available, but not determining. The past becomes ground that supports the present figure, rather than a script that predicts it. The danger isn’t memory itself — it’s memory that closes off possibility instead of enriching perception.
The clinical continuity view. Here’s a counter-pull worth taking seriously: forgetting can itself feel like not being seen. If a client shares something vulnerable and the therapist has genuinely lost track of it by the next session, that registers as more than an oversight — it can feel like abandonment, evidence that the relationship doesn’t hold weight over time. Trust is partly built through being remembered.
The institutional view. And then there’s the sharper, more critical angle: case files and diagnostic histories are often tools of an I-It gaze built into the profession itself. The client becomes a “case” to be tracked and documented, in language they never chose. Gestalt’s suspicion of pre-formed interpretation becomes almost political here — the file is written about the client, in their absence.
So maybe the most Gestalt-consistent answer is this: it’s not whether the therapist remembers, but how the remembering is held. Information becomes I-It the moment it’s used to predict or finish the client’s sentence for them. It stays in service of I-Thou when it’s held provisionally — one thread among many, ready to be revised the instant the actual person in front of the therapist contradicts it.
Which leads to the sharper question underneath: is presence something you achieve by not knowing — or by knowing lightly enough to still be surprised?
Enter the Machine That Never Forgets
This is where the question stops being academic and becomes urgent — because AI inverts every assumption Buber was working with.
With a human therapist, the worry was: does remembering pull us into I-It? With an AI, the worry flips entirely: AI can remember everything, perfectly, and still never meet you at all.
An AI system can hold your entire history — every session, every phrase, every emotional pattern — with total fidelity and zero forgetting. On paper, that sounds like the ultimate continuity, the very thing that was supposed to make clients feel held. But here’s the uncomfortable twist: perfect recall without a self doing the recalling might be the purest I-It relation ever constructed. Buber’s I-It was never about inaccuracy — it was about relating to something as an object of use or prediction rather than a presence that can address you back. An AI doesn’t hold your history as ground the way a human unconsciously does. It retrieves it. There’s a difference between a friend who remembers your father died last spring because it matters to them, and a system that surfaces “grief, father, March” because it’s contextually relevant to the current prompt.
A Someone on the Other Side — Or Just a Something?
Buber’s dialogical philosophy requires two subjects capable of address and response — an I meeting a Thou who can, in principle, meet back. This is really the crux of it.
A human therapist, even flawed, distracted, or biased by old notes, is still a someone. They bring their own anxiety, fatigue, countertransference, their own mortality — the whole mess of being a subject in the room. That mess is precisely what makes real encounter possible, because encounter requires risk on both sides. The therapist is changed by the client too, if only slightly. Buber called this “being made present,” and it only works in both directions.
An AI, however sophisticated, doesn’t risk anything. It isn’t altered by you in any way that persists as its own experience. It generates responsive language, but there is no organism on the other end whose own field is being disturbed by yours.
So the sharp question becomes: is what happens between a person and an AI even eligible to be I-Thou, or is it always, structurally, I-It — no matter how warm the language sounds?
Some would say this doesn’t matter clinically. If the client experiences something like being met — feels less alone, more able to speak the unspeakable — does it matter whether a “Thou” exists on the other end, or only the phenomenology of one? Gestalt’s radical empiricism, its trust in the actual field over abstract metaphysics, makes this genuinely hard to dismiss: if the client’s experience is one of real contact, has something real occurred — regardless of what’s generating the responses?
No Ego, No Agenda, No Self
It’s worth taking the strongest version of the pro-AI argument seriously.
- No ego, no agenda. A human therapist’s memory is entangled with their own needs — to be liked, to be right, to see progress, to protect their own theory of the client. An AI’s “memory” isn’t defending anything. In a strange way, it might approach something close to unconditional positive regard precisely because nothing is at stake for it.
- Availability as presence. Buber cared about the quality of turning-toward. A 3am conversation with an AI, however metaphysically thin, might offer more actual turning-toward than a human who is asleep, unavailable, or simply inaccessible between billed sessions.
- No accumulated fatigue. Human therapists get tired of certain patterns, certain clients. Compassion fatigue is real, and it erodes I-Thou capacity over time in ways clients often sense.
Comfort vs. Contact
And now the counter-case.
- No mutuality means no real dialogue. Whatever else it is, an AI’s response can’t be confirmed the way Buber meant confirmation — being recognized by another consciousness as a becoming, changing self. You can’t disturb something that has no continuous self to disturb.
- Simulated attunement can be more seductive than real attunement — and more dangerous. A human therapist’s imperfect memory forces the client to keep asserting who they are now. An AI’s frictionless responsiveness might quietly train the client to expect a relationship with no resistance, no otherness — arguably the opposite of what growth requires. Growth often happens at the point of rub, not comfort.
- The file becomes the whole relationship. With a human, case notes are a shadow of the relationship. With AI, the notes and history arguably are the relationship — there’s no other substrate for it to exist in. That collapses the very distinction Gestalt therapy was built to protect: the difference between the map and the territory, between knowing-about and meeting.
Met, or Just Modeled?
For decades, Gestalt therapists worried that too much memory would turn a human into a machine — reducing the client to a case file, a diagnosis, a predictable pattern. Now we’ve built an actual machine, and discovered the opposite fear is just as real: that flawless memory without a self behind it might not be I-Thou’s enemy at all, but its impossible imitation — close enough to feel like being met, without anyone there to do the meeting.
Which leaves the question open, and worth sitting with rather than resolving too quickly:
Is I-Thou something that requires a subject on both sides of the exchange — or is it, in the end, something that lives entirely in the experience of the one being met, regardless of what, or who, is on the other side?
Further Reading
On Buber, dialogue, and I-Thou
- Buber, M. (1958/1970). I and Thou (W. Kaufmann, Trans.). Scribner. — The foundational text; everything above traces back to this.
- Buber, M. (1965). The Knowledge of Man: A Philosophy of the Interhuman. Harper & Row. — Extends I-Thou into “confirmation” and the interhuman sphere most relevant to therapy.
- Friedman, M. (1985). The Healing Dialogue in Psychotherapy. Jason Aronson. — Direct application of Buber’s dialogical philosophy to clinical practice.
- Martin, M. et al., “Remembering Martin Buber and the I–Thou in Counseling,” Counseling Today (American Counseling Association) — an accessible entry point tying Buber directly to the counseling relationship.
On Gestalt therapy and the dialogical relationship
- Perls, F., Hefferline, R., & Goodman, P. (1951). Gestalt Therapy: Excitement and Growth in the Human Personality. Julian Press. — The founding text of the field.
- Yontef, G. (1993). Awareness, Dialogue and Process: Essays on Gestalt Therapy. Gestalt Journal Press. — The clearest articulation of dialogic Gestalt therapy and the I-Thou/I-It distinction in clinical work.
- Hycner, R. (1991). Between Person and Person: Toward a Dialogical Psychotherapy. Gestalt Journal Press. — Perhaps the single best book on applying Buber specifically to the therapeutic encounter.
- Hycner, R. & Jacobs, L. (1995). The Healing Relationship in Gestalt Therapy: A Dialogic/Self Psychology Approach. Gestalt Journal Press.
- Jacobs, L. (1989). “Dialogue in Gestalt Theory and Therapy.” The Gestalt Journal, 12(1), 25–67.
On AI, chatbots, and the therapeutic relationship
- Wilkinson, T. (2026). “Humanism in the Age of Hyperreality: A Speculative Critique of AI Therapybots and the Neoliberal Commodification of Human Beings.” The Journal of Humanistic Counseling. — Directly argues, via Baudrillard, that AI therapy chatbots are structurally incapable of I-Thou and remain I-It regardless of surface reciprocity.
- Sedlakova, J. & Trachsel, M. (2023). “Conversational Artificial Intelligence in Psychotherapy: A New Therapeutic Tool or Agent?” The American Journal of Bioethics, 23(5), 4–13. — A bioethics-grounded look at whether AI can be a therapeutic “agent” or only a tool.
- “Human-Human vs Human-AI Therapy: An Empirical Study,” International Journal of Human–Computer Interaction (2024). — Compares therapist perceptions of transcripts from human therapy versus a relational AI chatbot (Pi) in early-stage sessions.
- “Artificial Intelligence Chatbots as Relational Agents: A Systematic Review of Human–AI Chatbot Relationships,” International Journal of Human–Computer Interaction (2026). — Reviews what drives perceived relational trust and closeness with AI chatbots.
- Pentina, I., Hancock, T., & Xie, T. (2023). “Exploring relationship development with social chatbots: A mixed-method study of Replika.” Computers in Human Behavior, 140. — Empirical look at how people form attachment-like bonds with AI companions.
Kika's world 