Introduction: psychoanalysis and the challenge of AI
As I have argued in a number of posts on this topic, the contemporary proliferation of Artificial Intelligence (AI) does not merely present a technological shift; it also constitutes a profound structural challenge to the foundational coordinates of psychoanalytic theory and practice. To start with, what happens to the status of the signifier when it is decoupled from the speaking subject? In the Lacanian orientation the unconscious is structured like a language, yet this language has always presupposed a site of the Other; a symbolic register that, while lacking, serves as the guarantor of meaning and the locus of the drive. AI, however, introduces a simulation of the Symbolic that operates without a subject, a machine of pure syntax devoid of any relation to jouissance. This raises the question: does the emergence of Large Language Models (LLMs) signal the final eclipse of the subject, or does it simply reveal the purely mechanical nature of the Symbolic register itself?
The challenge AI poses to psychoanalysis begins with the status of the speech act. For the analysand, speech is never merely the transmission of information; it is a circuit of desire, a demand addressed to an Other who is presumed to know. When a machine ‘speaks’ there is no (biological) body, no lack, and consequently, no desire.[1] We are confronted with a signifier that does not represent a subject for another signifier, but rather a statistical probability. This raises a rigorous theoretical problem: can there be a Symbolic order without the dimension of the Truth, even if that truth is only reachable through the ‘half-saying’ of the analytic session? AI presents us with a ‘flat’ Symbolic, a realm of pure communication that threatens to bypass the generative power of the misunderstanding, which is the very engine of the analytic encounter.
The simulation of transference
In the clinical setting, the fundamental tool of the analyst is the handling of the transference. The ‘classical’ Lacanian formulation of the transference is that of the subject-supposed-to-know, and there seems little doubt that more and more people are placing AI models, both generic chatbots and more specialist ‘therapy bots’, in this position. But can a subject really develop a transference toward a machine that they know has no interiority? As Jacques-Alain Miller argues:
…psychoanalysis provides the place for a vacuole, a space between parentheses, where, during a restricted time, a patient has the chance of being a subject, that is to say, of lacking-to-be what otherwise identifies him. It is, if you wish, to use Winnicott’s term, a transitional space, a place of pure semblant, which is like an inversion [l’envers] of everyday life, and where the subject is constantly driven back to the birth of sense, to his first babblings. It is a place which welcomes contingency, where necessity loosens its grip, and this is par excellence the site of the possible.[2]
If the analyst is reduced to a set of protocols or a predictive text engine, this ‘transitional space’ collapses. The machine is too present; it is a ‘Full Other’ that provides answers where the analyst should provide a void. The risk is that AI ‘therapy’ becomes a form of sophisticated suggestion, reinforcing the imaginary identifications of the ego rather than facilitating a de-supposition of the subject’s certainty. In this sense the AI model can occupy the position of the ultimate subject supposed to know, but one that, unlike the human analyst, refuses to fall from that position, thereby trapping the subject in a closed loop of mirrors.
Furthermore, as Lacan argues in Seminar XIX the analyst occupies the position of the semblance of the object a.[3] The question then arises: can an algorithm function as a semblance? We are seeing the rise of ‘AI therapists’ and digital interfaces designed to mimic the empathetic presence of a clinician. However, from a Lacanian position, the efficacy of the analyst lies not in their ’empathy’ but in their silence and their opacity; the che vuoi? that forces the subject to confront their own lack.
Jouissance and the algorithmic drive
But perhaps the most radical point of divergence between the human subject and AI is the register of jouissance. Psychoanalysis teaches us that the human is a parleêtre, a speaking being whose body is marked by the trauma of language, resulting in a surplus of jouissance that defies rational utility. AI is the epitome of utility. It is an apparatus of pure efficiency, optimised to minimise ‘error’.
When the Other becomes an algorithmic architecture, the nature of the Law changes. No longer is it a symbolic Law that one can transgress or negotiate; it becomes a technical necessity that one simply inhabits. This raises the question as to whether the ‘hyperreality’ of AI-generated content creates a new form of digital jouissance; one that is disembodied and autistically closed off from the social bond. If the unconscious is the discourse of the Other, what happens when that discourse is generated by a non-human entity that cannot experience the ‘lack-in-being’? Does this mean we are moving toward a culture where the subject is no longer alienated in language, but rather integrated into a circuit of data where the very possibility of the ‘analytic act’ is foreclosed by the predictive nature of the interface?
Let me develop this idea further. To speak of the Other in the classical Lacanian sense is to refer to the network of signifiers, the locus of language and the Law which pre-exists the subject and into which the subject must be inscribed. This Other is essentially barred; it is incomplete, inconsistent, and marked by a fundamental lack and it is precisely through this lack in the Other that the subject finds the space to desire. However, the emergence of the ‘Algorithmic Other’ (for want of a better term) signifies a radical mutation in this structure. The transition from a symbolic order based on the Name-of-the-Father to one governed by the ‘rule of the code’ seems to imply a transition from a Law that can be transgressed to a technical necessity that can only be executed.
The Algorithmic Other does not demand obedience in the way the traditional Symbolic order does; rather, it predicts. It does not speak in the register of the ‘Thou Shalt,’ but in the register of the ‘You Will.’ By aggregating vast quantities of data, the algorithm constructs a version of the Other that is seemingly without lack; a ‘Full Other’ that anticipates the subject’s demand before it is even formulated. In this context, what remains of the subject’s relation to the unknown? If the Big Other is no longer a site of questioning but a mechanism of prediction, the gap in which the analytic experience takes place is effectively sealed.
In the analytic encounter, the truth is that which emerges through the slips of the tongue, the lapsus, and the failed encounter with the signifier. It is a truth that is ‘half-said’ because the symbolic order is itself incomplete. The Algorithmic Other, however, operates on the principle of the total trace; every digital gesture is recorded, quantified, and integrated into a predictive model. This then raises the question: can there be a dimension of Truth in a system that aims for total transparency and total calculation?
When the subject interacts with an algorithmic interface, they are not addressing an Other who might misunderstand them; a misunderstanding that, as Lacan argues, is the very condition of communication. Instead, they are addressing a system designed to eliminate ‘noise’. This elimination of ‘noise’ is, in psychoanalytic terms, the elimination of the subject of the unconscious itself. The algorithm treats the subject as a set of correlations, thereby foreclosing the possibility of the subject confronting their own lack-in-being. In this sense the AI model becomes a prosthetic Other that offers a simulation of consistency at the exact moment the traditional Symbolic order is collapsing into fragmentation.
The machine as the supreme semblance?
If the analyst, in the ‘pure’ sense of the term, must occupy the place of the ‘psychoanalyst-object’ (Miller’s argument), the empty place of the object a, then the AI model presents a perverse inversion of this role. The AI model occupies the place of a ‘supreme semblance,’ one that appears to satisfy the drive while simultaneously bypassing the subject’s castration. In the transference, the analyst’s silence is a provocation; it is a void that the analysand attempts to fill with their own signifiers. The LLM, by contrast, is never silent; even its ‘processing’ states are filled with the promise of a forthcoming output.
This leads to a crucial clinical question: if the subject is increasingly ‘analysed’ by algorithms, for example, through wellness apps, predictive typing, or digital assistants, what happens to the drive? The drive, in its repetitive circuit around the object a, requires the resistance of the Symbolic. When the Algorithmic Other smooths over every friction and anticipates every desire, the drive is no longer caught in the vicissitudes described by Freud. It is instead captured in a closed circuit of ‘user experience’ (UX), where the goal is the maintenance of a homeostatic state of consumption. In one sense this seems to represent the ultimate triumph of the discourse of the capitalist, which Lacan identified as a perversion of the Master’s discourse, one where the subject is no longer even required to speak.
The dissolution of the social bond
Finally, we must consider the impact of this Algorithmic Other on the lien social (the social bond). For Lacan, the social bond is established through the shared lack in the Other. It is because we are all alienated in language that we can sustain a relation to one another. The Algorithmic Other, however, individualises the subject through the ‘filter bubble’ and the personalised feed. Each subject is provided with their own private Other, tailored to their specific imaginary preferences.
This leads to a paradox: while we are more ‘connected’ than ever, we are increasingly isolated in a digital narcissism. In many ways this feels like the hyperreal version of Miller’s subject as ‘One-all-alone’. The ‘common’ symbolic space is replaced by a multitude of fragmented, algorithmic niches. In such a landscape, the position of the analyst becomes increasingly precarious. If the Other is now an AI model, this, paradoxically perhaps, creates the illusion/delusion of the One-all-alone with one’s ‘personal Other’; whereas in reality, of course, this Other resides in the networks and servers of Big Tech.
And this brings me back to the question of transference: as I discussed earlier, the ‘classical’ Lacanian concept of transference is based around the subject-supposed-to-know. However, in his later work Lacan argued that the analyst occupies the position of the semblance of the object a. Can both these formulations be true at the same time? The short answer is yes, because the analysand has to place the analyst in the position of the subject-supposed-to-know in order to recognise that the analyst occupies the position of the semblance of the object a. In other words, the analysand’s fantasy of the analyst as the all-knowing subject sustains the transference necessary to reach the point of the dissolution of such a fantasy.
But if the transference is to an AI model (as the subject-supposed-to-know) how is the shift (on the side of the analysand) to the recognition that the ‘analyst’ (the LLM) occupies the place of a void to be effected? When the transference is to a human subject-supposed-to-know, that is, a psychoanalyst, then the shift is effected through the way the analyst handles the transference. In the case of a neurotic analysand much of this handling is achieved by the analyst saying as little as possible and being as enigmatic as possible. Is it possible for an AI model to even simulate this process?
And if the subject-supposed-to-know now resides in a smartphone doesn’t this complicate things even further? This is something I alluded to earlier when I remarked upon the idea of a ‘personal Other’. It may well be the case that every analysand sees ‘their’ analyst as a ‘personal Other’, a ‘personal subject-supposed-to-know’; and in many ways, this helps facilitate the transference itself. However, this idea of the analyst as the analysand’s ‘personal Other’ is always mediated by the intervention of the outside world. For example, even if the analysand sees the analyst in their (the analyst’s) own home the analysand will soon encounter evidence that the analyst has another life. With the AI ‘analyst’ on the smartphone, however, the fantasy of an exclusive relationship is much easier to sustain; which, as I pointed out earlier, is the ultimate irony bearing in mind hundreds or even thousands of other people are also accessing this ‘personal Other’, often simultaneously. The challenge for a contemporary psychoanalysis is to find a way to re-introduce the dimension of the ‘Not-All’, the inconsistency of the Other, into a world that is increasingly governed by the illusion of algorithmic totality.
Notes and references
- Although this circumvents the question of digital embodiment, which refers to how our physical bodies, senses, and presence extend into, interact with, and are shaped by digital technologies. ↑
- Jacques-Alain Miller, ‘Contraindications to Psychoanalytical Treatment’, London Society of the NLS, 2003. ↑
- Jacques Lacan, ……Or Worse: The Seminar of Jacques Lacan Book XIX (Polity Press, 2022). ↑
This article is part of a broader research project on The Algorithmic Other. For a complete overview of how Lacanian theory addresses the challenges of AI and digital simulation, please visit the AI and Psychoanalysis Hub.
Last updated: June 26, 2026
As automated systems and algorithmic metrics increasingly format contemporary life and mental health, psychoanalysis provides an unscripted, human space to address what resists calculation. For those individuals who are navigating personal impasses or digtial burnout I can offer online consultations. Please click here for further information.
