A new collaborative study by computer and social scientists at the University of California, Riverside has sounded an alarm about the state of the internet. As people increasingly turn to artificial intelligence for quick answers, the researchers argue that the very soul of the internet—human emotion, lived experience, and messy, opinionated thinking—is quietly fading away. The study, which compared how AI systems like ChatGPT and Gemini respond to subjective questions versus traditional web searches, paints a stark picture of a digital landscape stripped of nuance and personality.
The researchers asked both AI chatbots and traditional search engines opinion-heavy questions, such as whether governments should ban fossil-fuel cars or whether the U.S. healthcare system needs reform. They then analyzed the reasoning behind each response using Aristotle's three pillars of persuasion: logos (logic and facts), ethos (authority and credibility), and pathos (emotion and human experience). Human-written blogs and search results typically blend all three appeals, creating rich, engaging content that reflects diverse perspectives. In contrast, the AI systems relied almost exclusively on logos—cold, factual, and devoid of emotional resonance or personal authority.
“What we found is that humans essentially use all three of those, whereas LLMs essentially only rely on logos,” said co-author Kevin Esterling, a professor of public policy and political science. “It's not like talking to a person at all. It's just a machine that's predicting what words ought to be said in response to a prompt.” This distinction has profound implications for how we consume information and form opinions. Without ethos and pathos, AI-generated content lacks the authenticity and emotional connection that make traditional web searches a vibrant marketplace of ideas.
Consider the example of searching for a margarita recipe. An AI might instantly provide a clean, competent list of ingredients and steps. But browsing a human-written cocktail blog, you might stumble upon the story that the margarita is named after the Spanish word for daisy and was accidentally created by an Irish bartender in Tijuana who grabbed the wrong bottle. Such anecdotes do more than entertain—they weave context, culture, and human ingenuity into the fabric of knowledge. These are the tidbits that enrich conversations and create memorable experiences. As AI summaries replace these narratives, society loses not only entertainment but also the subtle education that comes from diverse human storytelling.
The study highlights a growing trend: more people are bypassing traditional web searches in favor of AI-generated answers for topics ranging from health and politics to ethics and personal advice. While AI offers speed and convenience, it also filters out the messy, opinionated, and often contradictory voices that have long made the internet a dynamic repository of human thought. The researchers warn that this shift could gradually erode the public's exposure to diverse reasoning, leading to a more homogenized and less critical understanding of complex issues.
To fully appreciate the stakes, it's important to understand the methodology behind the study. The team selected subjective questions that inherently require moral reasoning and personal judgment—questions that have no single correct answer. They collected responses from multiple AI models and compared them to top results from traditional web searches and popular blogs. Using a coding framework based on Aristotle's three appeals, they scored each response for the presence of logos, ethos, and pathos. The results were consistent: AI responses were strongly skewed toward logos, while human-generated content showed a balanced use of all three appeals. This imbalance, the researchers argue, could have long-term consequences for how society debates and resolves contentious issues.
Historically, the internet has been a space where individuals share personal stories, challenge authority, and express emotions in ways that inform and persuade. Political debates, health forums, and hobbyist communities all thrive on the blend of factual data and human experience. The rise of AI as an intermediary threatens to flatten this landscape. For instance, in a discussion about healthcare reform, a human-written blog might cite personal anecdotes of medical debt alongside statistical evidence, creating a powerful case for change. An AI summary, by contrast, would likely present only the statistical arguments, missing the emotional weight that often drives public opinion and policy shifts.
The implications extend beyond mere entertainment. In areas like education, journalism, and civic discourse, the loss of pathos and ethos could lead to a more sterile, less persuasive exchange of ideas. Students relying on AI for research may miss the context that makes subjects come alive. Citizens seeking information on political candidates may miss the personal credibility indicators that help them assess trustworthiness. As AI-generated content becomes the default, the internet risks becoming a library of facts without the stories that give them meaning.
Some may argue that AI tools can be designed to incorporate more diverse reasoning. Indeed, recent efforts aim to fine-tune models to recognize and generate content that includes emotional and authoritative appeals. However, the UC Riverside study suggests that current models are fundamentally limited by their statistical prediction methods. They do not reason or feel; they calculate the most likely sequence of words based on training data. This inherent constraint means that even with better prompts, AI may never truly replicate the serendipitous and nuanced human interactions that make the web a vibrant space.
The researchers also note that the trend is self-reinforcing. As more content is generated by AI and consumed by users, the training data for future models will increasingly consist of AI-generated text, creating a feedback loop that further dilutes human perspective. This “data homogenization” could accelerate the loss of diverse voices, especially those from marginalized communities whose experiences are often best conveyed through emotion and personal narrative. The internet's soul might not be gone yet, but it is surely quietly fading, replaced by an efficient but hollow echo of human thought.
In response to these findings, the team urges users to be mindful of how they consume information. While AI tools are invaluable for certain tasks, they should not replace the rich tapestry of human-generated content. For now, the choice remains: we can let machines provide all the answers, or we can continue to seek out the stories, emotions, and arguments that have defined the internet since its inception. The research serves as a poignant reminder that the most interesting answers often come with a bit of personality attached.
Source: Digital Trends News