Artificial intelligence chatbots like ChatGPT are convenient, but a new study reveals they may lead to shallower knowledge retention compared to traditional web searches. Researchers at the University of Pennsylvania found that people who rely on LLMs (Large Language Models) to gather information develop less in-depth understanding than those who use search engines such as Google or Bing.
The Illusion of Effortless Knowledge
The core issue isn’t whether LLMs can provide answers—they can. It’s that the ease of these answers bypasses a crucial step in learning: synthesizing information from multiple sources. When you use a search engine, you encounter a range of perspectives, forcing you to evaluate, compare, and integrate ideas. LLMs, by contrast, deliver pre-packaged summaries that require less cognitive effort.
This matters because deep knowledge isn’t just about memorizing facts; it’s about understanding how those facts connect and why they matter. The study involved over 10,000 participants researching topics like gardening or healthy living. Those who used ChatGPT produced less informative summaries of what they learned—and were less likely to follow their own advice later.
The Link Problem: Even with Sources, Users Don’t Dig Deeper
The researchers even tested ChatGPT with built-in links to original sources. Surprisingly, only about 25% of users clicked on these links. This suggests that the convenience of the chatbot’s summary discourages further exploration. People aren’t just getting answers; they’re avoiding the work of deeper learning.
“LLMs can reduce the load of having to synthesize information for oneself… This ease comes at the cost of developing deeper knowledge on a topic.” – Shiri Melumad, University of Pennsylvania
The Solution Isn’t Avoiding AI—It’s Using It Wisely
The findings don’t mean LLMs are useless. As Carnegie Mellon psychologist Daniel Oppenheimer points out, users can still benefit if they follow the provided links or independently fact-check the information. The problem isn’t the tool itself, but how people naturally use it.
To learn effectively with AI, treat it as a starting point, not an endpoint. Click those links. Compare different sources. Challenge the chatbot’s responses. Only then can you unlock its potential without sacrificing the depth of your understanding.
The study underscores a fundamental truth: genuine learning requires effort. While chatbots offer convenience, true knowledge demands engagement, critical thinking, and the willingness to explore beyond the surface.
