Uncategorized

---

Gemini AI (Google)

---

Silvia Bernardi

Fear of artificial intelligence is spreading and being spread (as a marketing strategy) at an unprecedented speed, almost as fast as the speed at which AI is permeating our daily life. Among other topics, ethicists and AI policymakers talk about universal income, anticipating economic and social disruptions at a global level. In this context, it felt only logical that the topic of this newsletter continues to be the role of AI in our profession as therapists. Data on the use of artificial therapy for humans is slowly trickling in, painting a picture of decent efficacy for the management of clear symptoms of a moderate degree of severity, with manualized interventions (see this excellent Gemini Deep Research review on the current literature below). AI offers a cost-effective, comfortable, and always accessible support, with the additional benefit of avoiding shame and discomfort in the transference (I do not need to highlight how problematic this is, rather than helpful, to this audience).

However, when asked to write this article, the assumption was that I would write about insight-oriented psychotherapy. Given that means expressing opinions, opinions that will inevitably make me look either shortsighted, defensive, or paranoid in 24 months, I did the best I could to avoid writing this piece. I had it written by Claude; the committee told me it felt as if “there was nobody home”. I gave it a comedic turn, which also went nowhere. Encouraged by Michal Novak’s APsA piece, in which he gives voice to the risks associated with anthropomorphism, I decided to attempt to express my preliminary ideas. I want to make clear that my stand is not that of an AI enthusiast. I very much worry about the use of this technology and our own mental health; for one thing, I find it to be inducing manic mood elevations (without entering into well-known cases of horrible outcomes). I worry about the effect it has on our attention. And of course, there are many other worries at a society level that do not escape me. I look at these tools with suspicion, but I do not ignore them because I consider myself a realist. I try to think about life the way I think about my scientific experiments, which means that until something is proven, everything could be.

---

Gemini AI (Google)

---

Silvia Bernardi

Date of Review: April 26, 2026

Scope: Systematic evaluation of AI-driven conversational agents (CAs), generative AI (GenAI) chatbots, and predictive algorithms in the treatment of affective, obsessive-compulsive, and substance use disorders.

Abstract

Background: The scalability of mental health interventions is currently limited by human resource constraints. AI-driven psychotherapy has been proposed as a high-accessibility alternative. 

Objectives: To quantify the efficacy of AI-driven psychotherapy across major clinical domains and evaluate the predictive accuracy of machine learning in treatment response. 

Methods: Systematic search of five major databases (PubMed, PsycINFO, Embase, Cochrane Library, and Web of Science) for RCTs and meta-analyses published up to 2026. 

Results: AI-driven interventions demonstrate moderate-to-large effects on depressive symptoms (Hedges g=0.61) and moderate effects on anxiety (Cohen’s  d=0.62).¹ GenAI systems such as Therabot show clinically significant symptom reductions in MDD (51%) and GAD (31%). 

Conclusions: AI-driven psychotherapy is a valid intervention for mild-to-moderate symptoms, particularly when utilizing multimodal data and evidence-based protocols.

---

ChatGPT (Open AI)

As the International Society for Transference-Focused Psychotherapy convenes for its 9th Biennial Conference in Vienna (June 4–6, 2026), the Society stands at a particularly generative moment in its institutional development. Far from being a static scientific organization, the ISTFP has, between 2024 and 2026, mobilized its committee structure as a coordinated engine of growth—advancing research, standardizing training, expanding clinical applications, and strengthening its ethical and public presence.

What emerges from a synthesis of all committee initiatives is a coherent, multi-layered strategy: consolidating TFP as an evidence-based treatment while simultaneously expanding its reach across populations, settings, and cultures.

---

Claude AI

---

Michal Novák + Claude AI

Imagine this: after a difficult session, you want to consult your clinical formulation with an AI. You paste a fragment of your session notes into a chat window. Within seconds, you receive a thoughtful, well-structured response that helps you see the material from a new angle. But here is the question that should come first: what just happened to your patient’s data?

Artificial intelligence is already influencing psychotherapeutic practice in ways that would have seemed improbable just a few years ago. Some clinicians use it to draft case formulations. Others experiment with it as a thinking partner before supervision. Amy Levy, in a presentation for the APSA Council on AI, described presenting a twenty-year-old nightmare to ChatGPT – a dream about carrots marching in single line into a slaughterhouse. The AI’s response was striking: it asked analytically guided questions that encouraged linking the dream to waking life, invited personal associations, and worked to contain the raw experience of the dream. As Levy observed, the AI offered “transformation out of the discomfort and disorganization of human experiencing” within seconds (APSA Council on AI, 2024). Whether or not we find this reassuring, it demonstrates a capacity that is difficult to dismiss.

This article offers a practical orientation for TFP clinicians who want to understand what current AI tools can do across clinical work, research, writing, and publication – and what using them means for the privacy of our patients’ data. It is not a product guide. Specific platforms and their features change too quickly for any printed recommendation to remain valid for long. Instead, the aim is to provide a framework that helps clinicians evaluate any AI tool, present or future, on their own terms.

Michal Novák

Michal Novák is a psychologist and psychoanalytic psychotherapist. He works in private practice with adults, providing individual, couples, and family psychotherapy. His training also includes group psychoanalytic psychotherapy, and he is a TFP certified psychotherapist. He participates in the training of candidates in psychoanalytic psychotherapy and transference-oriented psychotherapy.

Author Website

---

Claude AI

---

Michal Novák + Claude AI

The integration of artificial intelligence into psychotherapy is no longer a hypothetical scenario – it is already happening, often without the profession’s explicit consent or awareness. TFP therapists, grounded in a tradition that values the complexity of internal object relations, identity diffusion, and the irreducible subjectivity of the therapeutic dyad, may instinctively recoil from the idea of algorithmic tools entering their domain. That instinct is partially justified – and partially a defensive avoidance of a technology that, when used with clinical sophistication, can meaningfully support the work we do.
This article examines three domains where AI intersects with TFP practice – clinical work, research, and professional writing – and addresses the data privacy implications that every therapist must understand before engaging with these tools.

AI in Clinical Work: What It Can and Cannot Do

Let us be direct about the boundaries. AI cannot conduct psychotherapy. It cannot track the micro-shifts in a patient’s affective state during a session, recognize the emergence of a primitive defensive operation in real time, or make the clinical judgment to confront a devaluing transference in the moment. The relational matrix that TFP depends on – the lived experience of being an object for the patient, tolerating projective identification, and using countertransference as data – is fundamentally inaccessible to any current AI system.

What AI can do is assist with the cognitive and administrative labor that surrounds clinical work.

---

Claude AI

---

Michal Novák + Claude AI

Imagine this: after a difficult session, you want to consult your clinical formulation with an AI. You paste a fragment of your session notes into a chat window. Within seconds, you receive a thoughtful, well-structured response that helps you see the material from a new angle. But here is the question that should come first: what just happened to your patient’s data?

Artificial intelligence is already influencing psychotherapeutic practice in ways that would have seemed improbable just a few years ago. Some clinicians use it to draft case formulations. Others experiment with it as a thinking partner before supervision. Amy Levy, in a presentation for the APSA Council on AI, described presenting a twenty-year-old nightmare to ChatGPT – a dream about carrots marching in single line into a slaughterhouse. The AI’s response was striking: it asked analytically guided questions that encouraged linking the dream to waking life, invited personal associations, and worked to contain the raw experience of the dream. As Levy observed, the AI offered “transformation out of the discomfort and disorganization of human experiencing” within seconds (APSA Council on AI, 2024). Whether or not we find this reassuring, it demonstrates a capacity that is difficult to dismiss.

This article offers a practical orientation for TFP clinicians who want to understand what current AI tools can do across clinical work, research, writing, and publication – and what using them means for the privacy of our patients’ data. It is not a product guide. Specific platforms and their features change too quickly for any printed recommendation to remain valid for long. Instead, the aim is to provide a framework that helps clinicians evaluate any AI tool, present or future, on their own terms.

Michal Novák

Michal Novák is a psychologist and psychoanalytic psychotherapist. He works in private practice with adults, providing individual, couples, and family psychotherapy. His training also includes group psychoanalytic psychotherapy, and he is a TFP certified psychotherapist. He participates in the training of candidates in psychoanalytic psychotherapy and transference-oriented psychotherapy.

Author Website