Shorter and Colder: How AI is Changing the Way We Speak to Each Other

Do We Talk to People Like Machines?

Imagine a typical morning. A man walks into the kitchen where his wife is busy at the sink. Instead of saying, “Honey, would you mind starting the coffee for me?” he barks: “Coffee. Black. Now.” He isn’t trying to be rude. He is simply speaking the way he has spoken to Google Assistant for months. The imperative mode has escaped his phone and taken a seat at the dining table. This example is not some dystopian fiction; it is a measurable trend.

In November 2025, the journal Spectrum of Engineering Sciences published a study by three researchers—Umair Paracha, Dr. Naeem Fatima, and Amna Ikram—focusing on how voice assistants alter human speech. They analyzed 500 recorded dialogues with Siri, Alexa, and Google Assistant from 200 participants. The resulting data is difficult to ignore.

The Mechanism of How AI Changes Language

The average length of a sentence directed at an assistant drops by 39.4 percent—from 14.2 to a mere 8.6 words. Subordinate clauses have nearly vanished, their frequency reduced by two-thirds. Most striking, however, is the shift in the nature of these utterances. Indirect requests—the cultural “could you…”—once made up 42 percent of interactions. Today, they account for only 12 percent. Meanwhile, direct commands have surged from 18 to 64 percent.

Paracha and his colleagues call this “anticipatory communication behavior.” Users simplify their syntax in advance, having learned that the machine cannot grasp subtlety. The authors state plainly that users “strategically modify speech acts to match the transactional nature of assistants, sacrificing the social and relational dimensions of conversation.” This is a clear example of how AI changes language by forcing us to mirror its limitations.

In other words: the machine isn’t learning to be more like us. We are adapting to be more readable to the machine. While this makes the technology easier to use, it fundamentally alters how we communicate, even when the technology is replaced by another human being.

Adopting the Persona of the Algorithm

Linguists from universities in Sweden and Denmark go even further. In a 2025 programmatic article published on arXiv, they describe a mechanism called “socio-indexing.” Accent, intonation, and style are not just tools for communication; they are signals of social identity. When we talk to someone for an extended period, we unconsciously begin to mirror them. The problem is that our most frequent conversational partner is now an algorithmically generated voice.

The authors note that “speakers may actively, though often unconsciously, incorporate features of AI language into their habitual speaking style.” A short-term adjustment, repeated often enough, leads to “long-term changes at both the individual and community levels.”

By designing assistant voices to be perfectly fluid and devoid of hesitation or pauses, tech corporations are silently defining a new communicative norm—without ever asking for our consent. Of course, as assistants get better at mimicking us, they might one day enrich our speech. For now, however, the scientific outlook is less optimistic.

Insights from Professor Jerzy Bralczyk

Professor Jerzy Bralczyk, one of Poland’s most prominent linguists, has spent years observing how language shifts under the pressure of external models. In one interview, he remarked:

The most credible people are those who speak spontaneously.

When we learn to formulate sentences dictated by an algorithm—short, command-based, and devoid of digression—we begin to speak with a heightened awareness of technique. We “try” too hard, and from there, it is only a small step toward pretense.

Bralczyk also warns of another danger: the “inflation of words.”

Every inflation breeds devaluation—if we talk about love too much, love becomes commonplace,

says the linguist. Generative algorithms are inflationary by nature; they produce words in excess, smoothly and without the weight of choice. Every sentence sounds confident. None of them carry a cost.

The Final Cost of Optimization

An optimistic interpretation suggests that language has always evolved alongside its tools. Writing changed speech, the telephone changed writing, and the SMS shortened our sentences. One could argue that “Coffee. Black. Now.” is simply a new register—functional, efficient, and pragmatic.

However, the data from Spectrum of Engineering Sciences reveals a frustrating reality: 82 percent of participants experienced issues with the assistant’s memory of previous conversations. We are learning to speak like robots, yet the robots still fail to understand us.

This leaves us with the question indirectly posed by Professor Bralczyk when describing speech as a space for Truth: is a “flawless” language—stripped of pauses, hesitations, and “unnecessary” words—still human speech? Or is it merely a transmission code: efficient, sterile, and ultimately hollow?

AI Cannot Grasp Emotion

If our effort to speak the language of machines robs us of our credibility, we lose more than just style. We lose what language did long before the invention of the alphabet: it built intimacy. Not through precision, but through imperfection.

Intimacy is found in the hesitation mid-sentence that says, “I am searching for words because I care that you understand.” It is in the digression that says, “I want to stay with you a moment longer.” It is in the question asked not for an answer, but to sustain a connection.

A voice assistant will never ask “how are you?” without a programmed reason. It will not leave a silence in which you might say more than you planned. It will not stumble over words out of emotion. It is precisely these imperfections—the pauses, the stumbles, the sentences cut short—that constitute the language of love, friendship, and presence. And these are exactly what we lose as we learn to speak more efficiently, more concisely, and more “to the point.” That is also how AI changes language: not by enriching human connection, but by thinning it out.

Read this article in Polish: Coraz krócej i chłodniej. AI zmienia sposób, w jaki mówimy do ludzi