Customer Service AI May Make People Talk More Like Machines

What happened

New research reported by Digital Information World suggests that people unconsciously adjust their own language when talking to customer service AI, mirroring the short, transactional phrasing typical of chatbots and voice assistants. Rather than the technology adapting to sound more human, the study finds the reverse dynamic at play: customers begin to sound more like machines the longer they interact with one.

The research points to a linguistic mirroring effect, where people interacting with automated agents shift towards clipped, functional sentences and drop the conversational cues — pleasantries, elaboration, emotional nuance — they would typically use with a human agent.

Why it matters

For experience leaders, this is a signal that the design of conversational AI doesn't just shape efficiency metrics — it actively reshapes how customers communicate, and potentially how they feel during and after an interaction. If terse bot language trains customers into terser, flatter exchanges, that could affect satisfaction scores, sentiment analysis accuracy, and even how issues get escalated or resolved.

It also has implications for AI training data and quality assurance: if customer inputs become increasingly mechanical in response to bot tone, organisations risk a feedback loop where systems are trained on impoverished language patterns, making it harder to detect nuance, frustration, or edge cases that require human judgement.

The Renascence take

Most conversations about conversational AI focus on whether the bot sounds human enough. This research flips that lens uncomfortably: the more relevant question may be what humans are starting to sound like.

Tone in service design is never neutral — it's a behavioural cue that customers absorb and reflect back, often without noticing. A chatbot scripted purely for efficiency doesn't just process a query faster; it quietly resets the customer's expectations for warmth, patience and detail in every future interaction, including with human agents. Operators obsessed with containment rates should ask whether they're optimising the bot at the expense of the emotional literacy of their entire service channel. The fix isn't necessarily "make the AI warmer" — it's designing deliberate moments where tone signals that nuance and frustration are still welcome, so customers don't quietly learn to stop bothering to explain themselves.