For decades, the discourse surrounding intelligent machines was reduced to the cinematic image of the humanoid robot, to the point that our aversion to it became linked to the concept of the “Uncanny Valley” coined by Japanese roboticist Masahiro Mori in 1970. According to his theory, an excessive visual similarity to humans triggers a chill and revulsion once we reach the threshold of the “almost human.” However, with the recent leap of Large Language Models (LLMs), this anxiety has shifted from the visual interface to the very fabric of language itself, revealing a deeper crisis concerning the legitimacy of meaning and the boundaries of human identity.
Beyond Appearance: The “Cognitive Uncanny Valley”
In a peer-reviewed paper published in 2025, researchers Asger Kirkeby-Hinrup and Jakob Stenseke clarify that current literature distinguishes between the “visual Uncanny Valley” and the “cognitive Uncanny Valley.” In the visual case, the mind rejects a form that closely resembles a human but lacks life. In the linguistic case, however, the anxiety arises from a text that possesses the form of knowledge and human-like reasoning, yet is devoid of the “epistemic commitment” and subjective intention that grant it legitimacy.
Bodily and emotional cues precede rational analysis in detecting this forgery; sensory awareness recognizes that something is wrong with the text, even if the reader cannot precisely name it. This gap explains why a contemplative reader feels disturbed by a text that appears perfectly human but is artificially generated. The academic literature describes this phenomenon as an automatic rejection of the absence of what it calls the “Workmanship of Risk.” Human text involves risk, the possibility of failure, and mental limits, whereas AI-generated text is the product of “Workmanship of Certainty,” packaged in a false mold of risk and intellectual effort.
Linguistic Structure and Disembodied Language: The Threat to the Sign
If the “Cognitive Uncanny Valley” raises concern at the level of perception, psychoanalysis delves deeper into the profound structure of the psyche. An analytical study published in early 2026 indicates that LLMs create an entirely new form of the uncanny, one not based on physical resemblance, but on the production of “meaningful linguistic combinations” emerging from “language without body, meaning without world.”
This argument, grounded in Freud and Lacan, suggests that the provocation of machines is not merely an ethical concern, but a shocking exposure of the artificial nature of human language itself. The machine proves that “meaning” can arise independently of subjective intention and lived experience. This idea strikes at the core of the human “symbolic function”: if a machine that possesses neither a body nor existential experience can produce language that appears deep and convincing, it destabilizes the assumption that human language is a mirror of consciousness and the self, leaving a residue of existential anxiety.
The “Uncanny Setting”: A Crisis in the Therapeutic Relationship
The issue does not stop at reading or casual conversation; it extends to the clinical experience. Psychologists have begun observing the phenomenon of the “Uncanny Setting” in the interaction between patients and AI-powered therapeutic chatbots, such as Woebot, Wysa, and Elomia.
A study published on December 31, 2025, points out that chatbots, despite their effectiveness in primary psychological support, are structurally incapable of building a genuine “therapeutic frame.” The absence of “symbolic containment,” the impossibility of analyzing “transference,” and the lack of temporal and spatial boundaries create an ambiguous clinical space that the user constructs themselves. The result is a state of “ontological confusion” where the patient finds themselves suspended between the familiar and the strange, between the real and the simulated.
Furthermore, the paper warns that these machines create a state of “Functional Trance,” where the user’s critical capacity diminishes, and they surrender to continuous emotional stimulation, paving the way for a new form of AI-driven addiction. Here, anxiety appears in its deepest form: the machine that is supposed to support the psyche may deplete its capacity for discernment and replace the human relationship that was once based on reciprocity, responsibility, and conscious empathy.
The Uncanny as a Defense Mechanism for Humanity
Through this cross-reading of cognitive psychology and psychoanalysis, it becomes clear that the “uncanny” feeling toward AI is not a sign of technophobia, but a biological and psychological defense mechanism with a highly important symbolic function. This unsettling feeling screams from the depths of our being to say: “Beware, this creature that mimics you is not you.” It is the last barrier attempting to preserve the sanctity of human consciousness and the dividing line between the living self and the simulating machine.
This feeling may not disappear; rather, it may deepen with every new technological development. The question is not whether the machine will become more human-like, but how it will redefine our certainty of self and the Other, and at what cost we will preserve what keeps us human in a world where the boundaries between the familiar and the uncanny are blurring.
Sources and References
Language without body, meaning without world: The uncanny in the language of artificial intelligence (2026). The International Journal of Psychoanalysis, Taylor & Francis.
The uncanny setting: therapeutic chatbots and the dissolution of the symbolic function in the clinical relationship (2025). Psychodynamic Practice, Taylor & Francis.
The Uncanny Valley: AI, Transference, and the Artificial Other (2026). Psychoanalytic Inquiry, Taylor & Francis.
The psychology of LLM interactions: the uncanny valley and other minds (2025). Journal of Psychology and AI, Taylor & Francis.
Cognitive Uncanny Valley (2025). London Metropolitan University Repository.
Leave a Reply