The Linguist

30 The Linguist Vol/64 No/1 ciol.org.uk/thelinguist OPINION & COMMENT We ask how GenAI is affecting our speech as robotic 'language tics' increasingly enter human communication In today's digital world, where Generative AI has become part of our daily lives, the line between human and computer-generated language is harder to draw than ever. Experts like Jess Hohenstein suggest that the more we interact with AI-driven tools such as chatbots, voice-recognition assistants, language translation apps and smart speakers, the more we begin to adopt the same linguistic patterns and structures they generate. 1 These systems are no longer just tools; they subtly influence the way we express ourselves. They shape our choice of words and our tone in ways we might not even notice. With every interaction, it feels as if we're gradually picking up 'language tics' from our digital companions. When applied to linguistics and technology the word 'tic' refers to recurring expressions or linguistic patterns that emerge in communication, particularly when interacting with AI systems like ChatGPT. These tics are words or phrases that AI systems tend to repeat based on their training data and text generation patterns. 2 The problem arises when people – whether consciously or unconsciously – begin adopting them thinking they sound more sophisticated, when in fact they are robotic and awkward. We refer to clichéd constructions commonly found in AI-generated responses ('I'm happy to assist you'), formal structures ('in the realm of', 'a tapestry of') and fillers ('in light of'). AI models often employ words like 'intricate', 'commendable' and 'pivotal' in ways that lack spontaneity and nuance, along with an overuse of adverbs such as 'methodically', 'strategically' and 'meticulously'. 3 As people mimic this form of communication we risk creating a kind of 'echo chamber' of empty expressions, contributing to a more mechanical way of interacting. This is what some researchers are already observing. 4 People gradually integrate these phrases into their speech without realising. Although similar language shifts have occurred in the past, with TV, books and internet slang shaping language, the influence of AI-generated content may signal the start of a new trend in communication. Research into AI-driven change While the influence of AI on written discourse has received considerable attention in the past two years, particularly regarding academic and scientific publications, research on its impact on spoken language remains scarce. However, it is clear that its influence extends across languages, with common phrases emerging in everyday speech, as seen in English ('navigating', 'unveiling', 'unlocking'), French (se plonger dans, embarquer, révolutionner, vibrant, notamment) and Italian (fatto luce, profonda comprensione, approfondimento), to name a few. Some studies have begun to explore this phenomenon. Ezequiel López and Iyad Rahwan analysed over 300,000 academic conference videos and developed a model to track the buzzwords generated by ChatGPT. 5 Their findings revealed that terms which were rarely used until recently, or had specific contextual meanings, have become commonplace in presentations and speeches. One such term is 'delve into'. The frequency of its use has led to a broader interpretation of the verb, making it interchangeable with others like 'look at', 'explore' or even 'start talking about'. From a scientific perspective, this shift in language use aligns with the concept of 'linguistic accommodation', a theory introduced by Howard Giles in the 1970s. The theory suggests that people adjust their speech to match the patterns and styles of those they interact with. When it comes to AI, this accommodation is reflected in people's tendency to attribute human-like agency and stereotypes to computers, as well as in the adoption of language tics. 6 In fact, evidence suggests that whether humans are communicating with other humans or technological agents, the effects of accommodation are similar. This raises thought-provoking questions about how such adaptations might influence our communication dynamics, merging human and machine language styles over time. Speak like a robot AI's influence could weaken cultural identity and lead to a homogeneous global language ecosystem CINTA GALLENT-TORRES RUBÉN COMAS-FORGAS

• Cover