The magic wears off! Why knowing more about AI makes you want it less
Here's a bet you'd probably lose. Picture two people, one can explain what a transformer is without mentioning Optimus Prime, the other thinks ChatGPT is a very fast intern living inside the internet. Which one is more eager to let AI write their emails, plan their holidays or diagnose their weird elbow pain?
Intuition says the expert, the data says the novice.
A study published in the Journal of Marketing (see link at the bottom) found exactly the opposite of what most of us, including the people they surveyed beforehand, would predict: the lower your AI literacy, the more receptive you are to using AI. The researchers ran four surveys, analyzed cross-country data, and conducted six additional studies, and the pattern held up stubbornly. People who understand how these systems actually work are, on average, less enthusiastic about handing tasks over to AI.
It's not fear. It's the death of magic.
The obvious explanation would be that knowledgeable people are scared, maybe they've read the papers or they know about hallucinations and bias. But the researchers ruled that out, the link wasn't explained by differences in how capable people thought AI was, how ethical they considered it, or how worried they were about AI ending humanity over a long weekend.
The real mechanism is more charming and a little sad: awe!
People with lower AI literacy are more likely to perceive AI as something close to magic. When a chatbot writes a heartfelt poem or seems to understand them, it feels like witnessing something impossible, a machine doing things that should require a soul or at least a pulse. That feeling of wonder pulls them in.
People with higher literacy don't get that hit. They know there's no ghost in the machine, just an enormous statistical model predicting the next token. And as the study's authors point out, it works just like a magic trick, the moment someone shows you the hidden compartment in the hat the rabbit stops being miraculous and starts being a rabbit. Knowing how the trick works doesn't make it bad, it just stops being magic. And apparently, nobody queues up to watch a magician explain his mortgage.
The "human" tasks are where it shows most
There's a telling detail in the findings. The literacy effect was strongest for tasks we associate with distinctly human qualities, things like writing something emotional or offering comfort. Watching a machine pull off something that feels human is precisely where the awe! kicks in for low literacy users, and precisely where high literacy users shrug and think "nice autocomplete."
For colder, more mechanical tasks like crunching numbers or analyzing data the gap shrank or even reversed. Nobody perceives a spreadsheet calculation as sorcery, no matter how little they know about computers.
Why this is awkward for basically everyone
This finding puts a fork in some comfortable assumptions.
For the tech industry, the default growth strategy of "educate people about AI and adoption will follow" might quietly backfire. The study suggests that boosting AI literacy could actually dampen enthusiasm, because education is functionally a magic-killing machine. The customers most eager to adopt your product may be the ones who understand it least, which is a genuinely uncomfortable sentence to put in an investor deck.
For educators and policymakers, we still absolutely should teach AI literacy. Informed users make better decisions, spot errors and are harder to mislead. But we should do it knowing the side effect. A population that understands AI may also be a population that's more selective, more skeptical and less impressed. That might be the goal, not the bug.
For the rest of us, it's a useful mirror. If you find yourself dazzled by an AI tool, it's worth asking whether you're impressed by what it does or by what you imagine it's doing. And if you're a jaded expert rolling your eyes at the hype, it's equally worth asking whether your indifference is insight or just the hangover that follows a revealed trick.
The takeaway
The study doesn't say ignorant people love AI and smart people hate it. It says something subtler: our openness to this technology is driven less by rational assessments of capability and risk, and more by an old human emotion, wonder, which fades with understanding.
Which leaves us with a strange truth about the AI era. The technology spreads fastest among those who understand it least, powered not by knowledge, but by the feeling that something impossible is happening.
The magicians knew this all along. The trick was never the hard part. The hard part is keeping people from looking behind the curtain.
Reference: Tully, S. M., Longoni, C., & Appel, G. (2025). Lower Artificial Intelligence Literacy Predicts Greater AI Receptivity. Journal of Marketing, 89(5), 1–20. doi.org/10.1177/00222429251314491