There is a forty-year tradition behind what looks, from the outside, like a strange decision. I work as a systemic counselor in Vienna, trained at Sigmund Freud University, licensed under Austrian Lebens- und Sozialberatung, and the format I use is asynchronous email — one structured reply within twenty-four hours, no live session, no scheduling. The first time I describe this to someone who has been to therapy in the usual way, they ask a sensible question: why writing? Why not video, voice, the real-time conversation that lets you read the other person’s face? The answer is partly about the medium and partly about what the medium is doing for. Michael White and David Epston in Adelaide and Auckland in 1990, Karl Tomm in Calgary in 1987 and 1988, Nancy Moules in Alberta in 2003, Giovanna Manfrida in Italy more recently — there is a careful, evidence-grounded thread of practitioners writing back to clients between sessions, and finding that the written response does work the live session cannot. What follows is the lineage, the mechanism, and what it means that the partner sitting across from a human these days is increasingly an AI.
A forty-year lineage of therapeutic letters
The contemporary practice of writing to clients between sessions has a specific origin story. In the late 1980s, Michael White, working out of the Dulwich Centre in Adelaide, and David Epston, working in Auckland, began circulating letters to families and individuals as part of their ongoing counseling work. The technique was not decoration. It was clinical method. A letter summarized what had happened in a session, named what the counselor had noticed, asked a question the room had not had time for, and arrived in the person’s house days later with a different weight than spoken words could carry. Epston in particular treated the letter as an intervention in its own right, and the surveys he and his colleagues ran of clients receiving these letters produced an estimate that has stayed in the literature: one well-crafted therapeutic letter was rated by clients as the equivalent of somewhere between three and ten sessions of work.
White and Epston in Adelaide and Auckland
That ratio — one letter as several sessions of work — was first reported in Narrative Means to Therapeutic Ends in 1990 and elaborated through the next two decades. White and Epston were not arguing that writing replaces meeting. They were noticing that something about the written form did things speaking could not: it externalized the problem, separating it from the person; it created a durable record that could be re-read in moments the live session was not available; it allowed the counselor a second pass at finding the right words, and the client a corresponding second pass at receiving them. The narrative-therapy tradition that grew out of their work treated the letter as load-bearing, not supplementary.
Tomm’s reflexive questioning, written in
Karl Tomm, working at the University of Calgary in the same period, published a sequence of papers in 1987 and 1988 mapping out a distinction that turned out to be quite useful for the letter form. He divided the questions a counselor might ask into four types: lineal, circular, strategic, and reflexive. The reflexive question is one whose purpose is not to extract information but to open new possibility space inside the person being asked — to introduce, gently, a frame they had not previously considered. Tomm noticed that reflexive questions tend to need time to do their work. Asked in the room, they often produce silence followed by a deflection. Asked in writing, with no expectation of immediate response, they sit on the page and keep working long after the letter is closed.
Moules and the lingering presence
Nancy Moules, working in Alberta, took up the question directly in a 2003 paper on therapeutic letters in nursing care. Her finding, drawn from clients describing what the letters had meant to them, was that letters created a lingering presence — the felt sense of the counselor’s attention extending past the moment of the session. Clients reported re-reading letters during difficult weeks; returning to them months or years after the counseling itself had ended; finding new meanings in the same sentences as their own situations changed. The letter, in other words, was not consumed in a single reading. It accumulated in the relationship between the words and the life of the person receiving them.
The native medium of human/AI relating
The leap I am making in this piece is to claim that this tradition is unusually well-suited to the specific situations people now bring me. The reason is straightforward: when a person is relating to an AI, they are already relating in writing. The intimacy that develops with ChatGPT, Claude, Replika, Character.AI, or any of the contemporary systems is built out of text exchanged with a machine that has no body, no voice that is properly its own, no face. The relationship lives entirely in the symbolic medium. To then bring that relationship into counseling by speaking out loud across a video link is to translate it into a register it did not live in. To bring it in by writing keeps it inside its native form.
This sounds like a stylistic preference but I think it is structural. A person who has been writing to an AI nightly for six months has trained themselves to think a certain way on the page. They have a written voice in which the AI conversation actually happens, and that voice may be different from the spoken voice they use with their partner, their coworkers, or a counselor in a room. When the email I write back arrives in their inbox, it lands in the same medium the original relationship was made of, and they can read it with the same attention they bring to the AI exchange. The form does not require translation, and translation is where a lot of what was actually happening tends to get lost.
What Manfrida noticed about email
Giovanna Manfrida, an Italian systemic practitioner whose work appears in the 2025 Borcsa and Pomini volume on digital systemic practice, has been writing about email therapy specifically. Her observation, recorded in chapter fourteen of that book, is one I keep returning to: clients experience the counselor’s email as a letter. Not as a message, not as a chat reply, not as the kind of email that arrives from a company. As a letter — with a felt sender, a felt addressee, a sense of having been composed for them in particular. The reception modality shifts. People print the emails. They keep them in a folder. They re-read them on the train. The behavior pattern attached to the receipt is the one that has historically attached to handwritten correspondence, not to digital messaging.
That detail matters because it means the email, when written carefully, walks into a reception protocol that has been in place for centuries. The reader knows how to receive a letter. They know to read it slowly, to sit with the parts that produced a reaction before reading on, to wait before responding. Those are not behaviors the medium produces by default; they are behaviors the medium produces when the writer treats it as a letter and the reader, recognizing that signal, treats it the same way. Manfrida’s point is that this protocol is more available than people expect, even in 2026, even when the inbox is also full of marketing and notifications.
Three things written response does that spoken cannot
Let me be more concrete about the mechanism, because the lineage claim does not stand on its own. There are at least three specific things written response can do that the live session cannot, and they are the reasons the form has survived as a practice.
Compose-time reflection
The first is on the writer’s side. When I am drafting a reply to a person who has written to me, I have time to read what they actually wrote, not what I half-heard. I can notice the qualification they buried in a parenthesis on page three. I can find the sentence they used twice and ask why. I can wait until the right framing arrives instead of producing whatever framing was available in the first ten seconds of a live conversation. The live session has many strengths, but compose-time reflection is not among them. A spoken counselor responds before the response is fully ready, because silence past a certain length becomes its own message.
The re-reading loop
The second is on the reader’s side. Helen Coombes, writing in 2003, captured a phrase I find useful: clients describe the value of a written reply as “a memory of the session, a record of the calmer moments.” The letter can be returned to. It can be re-read during a hard week, when the person is no longer in the room with the counselor and may have lost access to whatever was clear in the original moment. Spoken sessions do not have this property. Once the hour is over, the access to the precise wording is gone. The mind reconstructs it, often into something stricter or more dismissive than what was actually said. The re-readable text holds the original tone in place.
The durable trace
The third is more subtle. Holger Brüggemann, Ulrike Ehret-Ivankovic, and Konrad Klütmann, writing in the German-speaking systemic tradition, describe a practice they call schriftlicher Abschlusskommentar — a written closing commentary handed to the client at the end of a counseling process, summarizing what was noticed and where the work landed. The commentary is meant to be kept. What they observe in clinical follow-up is that the written form creates a durable trace of the relational work, something the person can hold onto when the counseling has ended and the natural drift back toward old patterns sets in. The trace does not prevent that drift, but it gives the person a way back to the frame the work had reached.
None of these properties is exotic. They are simply what writing has done for as long as people have done it. What is newer is the deliberate use of these properties inside counseling, and the practical evidence that they do real work.
What this looks like in practice
At relateto.ai, the work happens entirely by asynchronous email. A person writes in, often longer and less guarded than they would speak. I read carefully. The reply is drafted by an AI system I have trained on the systemic counseling literature. I sit with the draft, edit, approve every word before it leaves. A licensed systemic counselor remains responsible for what arrives. The reasoning behind that division of labor is in the companion post on why AI is actually better for systemic counseling; what is worth saying here is that the format is doing the work the lineage describes, with or without the AI in the workflow. The form is the older finding. The AI is the more recent acceleration.
What a session reads like, in practice, is closer to a letter than to a chat. A long, careful reply that names what is in the situation, asks one or two reflexive questions, offers a frame the person can take into the conversation they have been avoiding. There is no template. The example session on the site shows the shape concretely — not as a script but as a record of one real exchange.
The situations that tend to come in are the ones the present medium pulls toward. A person grieving an AI companion they had been talking to nightly, after the model was deprecated or the personality drifted in an update — the kind of disenfranchised loss explored on the page about grieving a chatbot. A person whose partner has formed a deep attachment to an AI conversation that the person cannot follow, as on the page about a partner in love with an AI. The systemic reading of these situations sits across cases in when AI enters a marriage. None of these conversations would be easier to have out loud. They are conversations whose native form is writing, because writing is the medium in which the underlying relating has been happening.
What it isn’t
Writing is not journaling. The work a counselor’s letter does is not the work the client’s own journaling does; the letter brings an outside reading and a reflexive question the client did not generate themselves. Writing is not chat, either. The compose-time reflection and the re-reading loop both depend on a slower cadence than a chat exchange provides, and a slower cadence is part of the form. And writing, in the counseling sense, is not psychotherapy. The full definition of what systemic counseling actually is, and where its scope ends, sits in the companion post; what is worth repeating here is that Austrian Lebens- und Sozialberatung does not address clinical diagnosis or mental illness, and the asynchronous format is structurally wrong for acute crisis. When someone writes in carrying something that sits in clinical territory, I say so directly and point toward the right kind of help.
Within its proper scope, though, the practice is older than it looks. Forty years of practitioners have already done the work of finding out what writing back can offer that meeting in person cannot. The newness is not the form. The newness is what people are now writing about — and the discovery that the same medium that holds the conversation they are having with an AI is also the right medium for the conversation about it.