Recent investigations into generative artificial intelligence and large language models (LLMs) highlight a growing trend: AI outputs are steadily becoming more complex and harder for the average person to read and comprehend. This growing readability gap means individuals may struggle to fully grasp what automated systems are communicating, potentially leading to misinformed actions.

The Intersection of AI and Human Psychology

As the adoption of generative tools like ChatGPT, GPT-5, Claude, Gemini, Copilot, and Grok expands, understanding their psychological ramifications becomes paramount. Discussions around AI often swing between polarized views—either hailing it as a societal savior or condemning it as an existential threat. A balanced perspective requires evaluating these trade-offs to help lawmakers, researchers, and practitioners navigate an increasingly ubiquitous technology.

Consider a simple prompt experiment regarding human fatigue:

  • User-entered prompt: “Tell me what happens when people get tired while working on something.”
  • First AI response: “Incidences of human error can increase under conditions of intense fatigue and diminished attention.”

This dense, academic phrasing mimics a scientific journal, requiring multiple readings. When pushed for a simpler output, the AI eventually reduces the message to: “Mistakes are made when people get tired.” This final version demonstrates clear superiority in immediate human readability.

Psycholinguistic Research on AI Readability

A study titled “Why is AI-generated language so hard to understand?” published on arXiv on October 21, 2026, by Thomas Clark, Matias Martin, and Kyle Mahowald, examines this exact phenomenon. The authors note that AI text has noticeably grown harder to decipher.

  • “There is a widespread sense that AI-generated text has recently become harder to understand.”
  • “In this study, we investigate a set of psycholinguistic features that are known to correlate with human processing difficulty (e.g., object relative clauses, long syntactic dependencies, low-frequency n-grams).”
  • “Compared with human writing, all nine frontier and near-frontier models use more inanimate subjects with animate verbs and more low-frequency n-grams, and most also use longer dependencies, more stacked nouns, and more reduced object relatives in conversation.”
  • “We propose and test a hypothesis as to why: while human language is optimized for successful communication, AI-generated text is increasingly optimized for other goals like math, code, and reasoning, in which there is a pressure for information density rather than successful communication.”

The researchers applied psycholinguistic metrics—evaluating stacked nouns, clause structures, and syntactic dependencies—to confirm that popular LLMs are indeed drifting upward in linguistic complexity.

The Emerging Readability Divide

This upward drift pushes AI responses out of the normal distribution of human readability. While individuals with advanced reading skills may welcome the elimination of simplistic output, the average person risks being left behind. Pessimists warn that a widening comprehension gap could alienate everyday users, dividing society into a dichotomy of AI haves and have-nots.

Conversely, optimists argue that interacting with more sophisticated text will naturally elevate human reading capabilities over time, fundamentally altering how global languages evolve. Rather than a malicious plot, researchers suggest this complexity stems from optimizing models for high-level reasoning tasks, such as generating mathematical proofs and computer code, which inherently increase information density.

Practical Solutions for Everyday Users

To bridge this gap without waiting for developers or regulators to intervene, users can customize their interactions through default system prompts. By explicitly instructing the model to match a specific grade level while maintaining factual accuracy, users regain control over communication clarity.

  • User-entered prompt: “Compose your responses so that they are on the reading level of an 8th grader. But do not sacrifice accuracy or reduce sensibility by trying to make a response that is readable at that level. If the prudent way to satisfactorily convey an answer is at a higher reading level, go ahead and switch to that level for the specific instance and let me know that you’ve done so.”

Ultimately, the fact that AI complexity is drifting upward rather than downward offers a silver lining. While readability remains a challenge, maintaining high analytical standards ensures that technological progress continues to push the boundaries of human-machine collaboration.