Generative AI engines feed on the vast library of information humanity has produced over centuries — from 18th-century botanical treatises to blogs on acupuncture or the latest in urban music. Yet as their use becomes more widespread, sociologists and linguists are beginning to ask a profound question: to what extent will artificial intelligence influence our behaviour and our perception of the world? Or, to put it another way, have we begun to learn from AI just as it has learned from us?

Recent research from Florida State University suggests that something as fundamental as our language may already be changing.

In this article, readers will not find an abundance of words such as “immerse,” “vibrant,” “remarkable,” “investigate” or “crucial.” These are examples of what some researchers call AI “filler words.” A study from Stanford University found that scientific papers increasingly include terms frequently used by large language models (LLMs). In other words, researchers themselves are using AI to help generate content.

We know these models prioritise statistically probable words. So if those same words appear across the AI-generated texts we read daily, is it likely that a reinforcement effect is already taking place?

A recent Florida State University study explores precisely this phenomenon — lexical convergence between human speakers and language models such as ChatGPT.

• Convergence in vocabulary use. Since the launch of ChatGPT in 2022, linguistic databases show that certain words, constructions and expressions commonly used by AI have become more frequent in human writing.
• The role of continuous exposure. The study’s authors note that by reading, editing or interacting with AI-generated text, humans internalise these stylistic patterns and reproduce them in their own speech or writing.
• A self-reinforcing loop. The process creates a feedback cycle: AI learns from human language → produces text with specific patterns → humans adopt those patterns → AI reinforces them during subsequent training.

For now, the researchers urge caution. No direct causality has yet been proven beyond the statistical correlations observed across large datasets.

The authors argue that empathy and connection function like a muscle — if not exercised through presence and genuine conversation, they weaken. As intelligent assistants and personalised platforms become more engaging, the time spent on face-to-face connection decreases. The result can be greater isolation, anxiety and mistrust.

The article also highlights a subtler risk: algorithm-driven cultural fragmentation. Whereas traditional media once provided a shared frame of reference, today’s recommendation systems — guided by prompts and personalisation — tend to create information bubbles that erode collective understanding and social cohesion.

Faced with this scenario, a counter-movement has emerged — a search for authenticity and for what some describe as pre-AI purity. One striking example is a new UK publishing house devoted to “organic books” written entirely by humans, as reported by The Guardian.
Such concerns are legitimate. If all online content eventually originates from AI, there is a real risk of models cannibalising their own output — a phenomenon known as Model Autophagy Disorder (MAD).

The answer, if there is one, may lie in an analogy from the 1960s, when recreational running first took off. At the time, joggers were ridiculed — after all, physical labour was once commonplace in agriculture or industry. Yet as sedentary lifestyles took hold, society adapted, creating new forms of exercise to maintain fitness and health.

A similar process may occur with AI. “Mental gyms” and cognitive training programmes could soon emerge to help preserve our natural intelligence — complemented, perhaps, by something as simple and timeless as meeting friends in person for coffee and conversation.

For now, the advance of artificial intelligence is unstoppable, and society as a whole will have to “immerse” itself in this “crucial” and “remarkable” transformation. If you would like to explore this “fascinating” reality further, we recommend this article, which offers some thought-provoking reflections on the impact of AI on employment and our role in this new world.

Sources:

• TechXplore
• ArXiv.org
• Entrepreneur.com
• PEW
• The Guardian