Why Most People Are Using AI Wrong
The thinking skill that separates useful output from noise.
Posted May 2, 2026 | Reviewed by Margaret Foley
Some people get shockingly useful, nuanced responses from AI models such as ChatGPT, Claude, or Gemini. Meanwhile, others get generic, surface-level output and assume the tool is limited or even useless to them. They aren't using it "wrong" in a technical sense, but in a way that limits what AI can actually do.
Why does the same tool produce such different results depending on the user?
The difference isn’t technical skill, but rather how people are thinking about the system they’re interacting with.
What effective AI users are doing is a form of system perspective-taking , a kind of "as-if" theory of mind applied to a non-human system.
The Hidden Skill: System Perspective-Taking
To be clear, AI doesn’t have a mind, but people often interact with it as if it does. The users who get the most out of it use system perspective-taking to anticipate what the system needs in order to produce a useful response before they ever type the prompt. In practice, this means recognizing what’s missing and supplying it: relevant context, constraints, goals , and assumptions.
It also means understanding the system’s limitations. AI doesn’t have real-world grounding and generates responses based on probabilities. As a result, vague inputs tend to produce generic, high-likelihood outputs. Rather than treating it like a one-shot search engine, effective users implicitly simulate a back-and-forth dialogue, refining their inputs as if they were briefing a collaborator.
The difference becomes obvious in something as simple as asking for help organizing a project. A vague prompt like “help me with this project” tends to produce a generic, surface-level list. But a more specific prompt (i.e., what the project is, your goal, your timeline, your constraints, and what you’ve already done) leads to something far more structured and actually usable.
This pattern mirrors what we know from research on perspective-taking: Communication improves when we anticipate how others will interpret our message. The Computers Are Social Actors (CASA) paradigm shows that people automatically apply interpersonal social rules to machines that use natural language, perceiving them as social partners and adapting their behavior accordingly, even when they know the system is artificial. In this case, treating AI as a kind of “as-if” thinking partner can actually lead to more precise, effective use.
What Effective vs. Poor AI Use Actually Looks Like
Beyond lackluster prompts, another common behavior of poor AI use is accepting the first answer uncritically. This is where the danger is: When AI becomes a substitute for generating ideas or evaluating information, its use begins to erode critical thinking.
In contrast, more effective users approach AI with intention. They refine their prompts based on what the system returns, treating the interaction as iterative rather than one-and-done. They actively evaluate outputs, noting what’s useful, what’s missing, and what needs to be sharpened. Essentially, they use AI as a thinking partner rather than a replacement. It's important to note here that the best AI users treat responses as drafts, not final answers. This kind of interaction also requires cognitive flexibility: the ability to adapt, revise, and reframe based on new information.
The key distinction is not that AI works better for some people. It's that some people are better at drawing out what it can actually do.
Active Thinkers vs. Passive Consumers
Ultimately, the people who get the most out of AI differ not just in how they use it, but in how they see themselves. Those who identify as idea generators tend to have far more success than those who see themselves primarily as information gatherers. The users who benefit most are not relying on AI to think for them; they’re bringing thinking to it. They contribute judgment and taste, elements that the system cannot generate.
AI can amplify thinking, but it cannot truly originate in the same way that our brains do.
People who approach AI as passive consumers often receive generic output, while those who approach it as active thinkers use it to expand, refine, and sharpen their own cognition .
There are, however, limits to this approach. Over-anthropomorphizing AI can lead to overtrust. This becomes a problem when we begin to treat outputs as if they reflect intention or true understanding. The goal isn’t to believe that AI has a mind, but to use as-if reasoning strategically. When applied deliberately, this kind of perspective-taking can improve how we interact with the system, without losing sight of what it actually is.
Using AI Well Is a Cognitive Skill
People who are better at using AI are able to engage in perspective-taking and see themselves as active thinkers and idea generators. That’s the difference. Using AI effectively is not about developing better AI; it’s about being a better thinker.
Ironically, this means that the people who get the most out of AI are often the ones who need it the least
Unfortunately, as AI becomes more embedded in daily work, the skill gap may widen, not shrink, if it isn’t addressed. A 2025 mixed-methods study found a significant negative correlation between frequent AI tool use and critical thinking ability, mediated by increased cognitive offloading, with younger users showing the highest dependency and lowest critical thinking scores.
The encouraging part is that this can be developed. But it requires intention. As AI makes it increasingly easy to outsource effort, the challenge will be to stay mentally engaged: to question, refine, and direct rather than passively accept. If approached this way, AI doesn’t diminish our thinking, it sharpens it.
Lee, J.-E. R., Nass, C. I., Lee, J.-E. R., & Nass, C. I. (2010). Trust in Computers: The Computers-Are-Social-Actors (CASA) Paradigm and Trustworthiness Perception in Human-Computer Communication . IGI Global Scientific Publishing. https://doi.org/10.4018/978-1-61520-901-9.ch001
Gerlich, M. (2025). AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking. Societies , 15 (1), 6. https://doi.org/10.3390/soc15010006
Schulhoff, S., Ilie, M., Balepur, N., Kahadze, K., Liu, A., Si, C., Li, Y., Gupta, A., Han, H., Schulhoff, S., Dulepet, P. S., Vidyadhara, S., Ki, D., Agrawal, S., Pham, C., Kroiz, G., Li, F., Tao, H., Srivastava, A., … Resnik, P. (2025). The Prompt Report: A Systematic Survey of Prompt Engineering Techniques (arXiv:2406.06608). arXiv. https://doi.org/10.48550/arXiv.2406.06608
Share this post Facebook Bluesky Linkedin Email
There was a problem adding your email address. Please try again.
By submitting your information you agree to the Psychology Today Terms & Conditions and Privacy Policy
Elizabeth Mateer, Ph.D., is a neuropsychology fellow at Harvard Medical School exploring the intersection of brain science, creativity, and modern identity.
Get the help you need from a therapist near you–a FREE service from Psychology Today.
This article is part of the Bringwise Psychology Journal — daily insights on human behavior, mental health, and personal growth.