Artificial intelligence is gradually entering the world of theater, influencing everything from writing and translation to staging. However, the unique nature of dramatic texts—where rhythm, breath, orality, and subtext are closely tied to performance—reveals the current limitations of AI. Theatrical translators are now navigating a complex space, trying to find their role in a creative process that is becoming increasingly hybrid. Since the 1950s, artificial intelligence has grown rapidly, becoming a part of everyday life. Today, AI can mimic human cognitive processes more effectively than ever, thanks to powerful algorithms and access to vast amounts of data. As a result, it is now influencing various sectors, including theater, which is facing new technologies driven by computational power. However, this is not the first time that theater has embraced new technologies. From the rotating stages of ancient Greek theater to the use of projections in the 20th century, the arts have historically adopted emerging techniques that reshape the creative process. From the puppet to the algorithm: the ambitions of the machine The rise of robotics and computing has opened new paths for theatrical research. The idea of using artificial figures in theater is not new—Heinrich von Kleist, in his 1810 text On the Puppet Theater, suggested that a puppet could be an ideal interpreter of art. Similarly, Edward Gordon Craig, in his 1908 article The Actor and the Super-Puppet, proposed that a mechanical figure might be more suitable for theater than a human, who is prone to imperfections. Today, AI is reviving these ideas with new urgency. The word "robot," coined in 1920 by Karel Čapek, comes from the Slavic word "robota," meaning "work" or "drudgery." In this context, digital tools have become essential for the theater world. Initially, AI was used for technical aspects like sound and lighting, but it now influences pure creation, from writing to staging. Experiments are multiplying: in 2020, the THEaiTRE project at the Academy of Performing Arts in Prague aimed to create a play entirely generated by AI. In 2025, a Sorbonne troupe created a play in the pure style of Molière using AI. Dramaturge Annie Dorsen has also experimented with "algorithmic theater," using texts generated by algorithms without human actors, challenging the digital impact on the illusion of our choices. These experiments blur the lines between human and machine, sparking speculation about the future of theater. AI is no longer just replacing the actor's body; it is now challenging the mind and the text itself. The translation market facing the neural transition AI's impact on theater is most evident in the realm of translation. The rise of Neural Machine Translation (NMT) in the 2010s has transformed the field, allowing machines to provide translations with a natural quality by considering the overall context of texts. Generative AI is now reshaping the role of translators, who are at the forefront of these changes. While AI offers economic and logistical advantages—such as fast, low-cost translations for large volumes—it also raises concerns. Deep learning enables these tools to self-correct, making them more reliable. However, the threat of AI is not as clear-cut. The translation market can be divided into two sectors: an automated mass market and a human niche market. Translators in the mass market are heavily threatened, while those in the niche market, who provide advanced expertise, remain secure. These specialists offer an aesthetic rendering that, for now, remains beyond AI's reach. The specificity of the theatrical text: orality against the machine Theatrical translation belongs to this niche market. AI translates literal texts, but theater demands a literary and scenic text. AI often produces translations that sacrifice narrative coherence and emotional relationships. It lacks the sensitivity to language and the creativity of a human translator, which are crucial for conveying the rhythm and emotion of the stage. A grammatically perfect reply on screen may be unpronounceable on stage. Here, words go beyond mere formulation: subtext, metaphor, ambiguity, and poetry are often lost in a low-quality dramatic rendering. The theatrical translator is, by essence, a dramaturge. The impact of AI's failures is immediately noticeable when moving to the stage. Researchers of the THEaiTRE project had to admit that they could not make the AI understand that characters should be animated by their own objectives and motivations. Similarly, an experiment by the Young Vic in London resulted in the algorithm generating texts that reproduced racist and stereotypical biases found on the internet. What "makes theater" remains beyond the machine's capabilities, which lack internal criteria for "theatricality." This theatricality needs to be clearly defined. Historically, this concept has been theorized as a set of supertextual information. Roland Barthes defined theatricality as "theater minus the text," a thickness of signs and sensations built on the stage. He described theater as a "cybernetic machine" that emits messages simultaneously but with different rhythms. Patrice Pavis expanded on this, stating that theatricality includes elements beyond the purely scenic, emphasizing the indispensable aspects of all theatrical phenomena. The specificity of the theatrical text lies in its intention to be created and shared in front of an audience. This performativity gives it unique writing characteristics. For Patrice Pavis, orality is a preliminary part of the text's concretization on stage: speech is integrated into the "spectacular text," where the spectator receives linguistic, vocal, and bodily signs simultaneously. For researcher Marion Chénetier-Alev, it is necessary to distinguish vocalization from orality. Vocalization is the physical act of pronouncing the text, while orality belongs to the writing itself. It is the imprint of breathing, rhythm, and flesh "encoded" in the page, which the reader physically feels. Orality is the body of the words on the page, and vocalization is the moment when this body meets that of the actor. Following this work, the translation of the body of the text imposes four major constraints that an algorithm cannot resolve: 1. The constraint of the "physiological engine," which favors breath over grammar, even at the expense of syntax to recover the original pulmonary delivery. 2. The constraint of the "visual partition," which respects the spaces, typography, and punctuation (translating the silences). 3. The constraint of the "acoustic analogy," which seeks for sound textures and similar timbres from one language to another. 4. The constraint of the "rejection of familiarization," which refuses a normalized or smoothed orality. The translator is engaged in a physical as well as intellectual posture. AI translates words; the translator understands actions, rhythms, and performance intentions. The lines are transposed more than they are translated. The translator as a mediator and director What place remains for the theatrical translator between artificial intelligence and this theatricality devoted to translating? The translator is a mediator of cultures and imaginaries. He welcomes the source language without erasing it. Philosopher Barbara Cassin highlights the difficulty of the machine in the face of "untranslatables," terms whose meaning varies deeply from one culture to another (like the English "spleen" or the Portuguese "saudade") and which require human arbitration. Furthermore, if the theatrical translator must express gestures and a space, his work is akin to a first staging. As Antoine Vitez pointed out: "A great translation already contains its staging." Jacques Rancière conceives of theater as a generalized space of translation, from the author to the emancipated spectator, where each person translates in their own way what they perceive, linking the performance to their singular intellectual adventure. Toward a hybrid approach: AI as a clearing tool Thus, the question is not whether AI will replace human intelligence, but rather to determine how it can serve the translator. Condemning its use will not prevent its spread. If AI encounters obstacles in high literature, it nevertheless presents itself as an effective clearing tool at the first level of reflection. It favors initial drafts, serves as a dynamic dictionary, accelerates document research, or proposes sentence variations according to the instructions provided. It assists the translator, but a meticulous review by a human expert remains essential to correct its syntactic approximations or its "hallucinations." Artistic control remains the domain of the human. Some professionals choose to integrate AI in post-editing: they generate a first raw text by the machine, then intervene to refine the tone and inject a missing theatricality. However, this method presents the risk of formatting the translator's mind, which remains constrained by the (non-theatrical) structure proposed at the beginning. Conversely, other creators, such as poet Camille Bloomfield, prefer an inverse approach: perform first a traditional translation work to preserve their initial human impression, then use different AI as partners to open new lexical fields and propose variants, the human always making the final choice.