After reading a book, we say we “know” its contents. We “know” what we just searched for on Google. The report ChatGPT wrote for us, we think of as “my output.” But do we, really? Between knowing and the illusion of knowing, there is a chasm far deeper than we tend to think.
In psychology this phenomenon is called the “illusion of knowledge” — the cognitive error of confusing information you’ve encountered externally with something you originally knew yourself. The illusion grows more frequent, and deeper, the more technology advances.
In 2015, a Yale University team led by Matthew Fisher ran nine experiments and found that people who had just searched the internet for information rated their own knowledge significantly higher than people who hadn’t. The astonishing part is that the overestimation extended beyond the topic they had searched, to wholly unrelated topics. Someone who had used Google to look up “how a zipper works” also rated themselves as knowing more about, say, “why a cloudy night is warmer than a clear one.” When asked to pick a brain-activation image that resembled their own brain, the search group picked more highly activated ones.
Adrian Ward of the University of Texas dug deeper into the roots of the phenomenon. The core finding from his eight experiments is that people who used Google could no longer properly distinguish between external information and their own memory. They even predicted they would do better on future tests without internet access. Ward’s explanation cuts to the heart: “Searching for information on Google is too similar to searching one’s own memory. As a result, people lose the ability to distinguish where the internet’s knowledge ends and their own knowledge begins.”
What’s interesting is that the depth of the illusion shifts with how the information is accessed. In Ward’s experiments, artificially slowing down Google search made the overconfidence vanish; providing Wikipedia links directly also reduced the illusion. The faster and smoother the interface, the more the “this is external information” awareness dilutes. AI takes this one step further. ChatGPT delivers polished sentences as if the thoughts in your own head had been organized for you. In a 2025 experiment by a team at Aalto University, most ChatGPT users overestimated their own performance, and people more familiar with AI showed even greater overconfidence. Most users put a question to ChatGPT once, accepted the answer as-is, and never verified or pushed back.
The point, in the end, is this. The knowledge inside a book is not yours while you are reading. It becomes yours only when you wrestle with it, digest it, and rebuild it in your own words. The information on Google and the output from AI are, in essence, the same. To borrow Fisher’s phrasing, “having access to the internet creates the illusion that the knowledge inside you has grown.” In the AI age, this illusion can only deepen and widen. The real danger is not that technology lowers intelligence — it’s that technology makes us mistake not-knowing for knowing. Socrates’ “know thyself” gains a new meaning in the AI age. Knowing precisely what you know and what you don’t know — that is the most important ability left to humans in the AI age.