5 Common Misconceptions About AI in Neuropsychology

Unless you’ve been living under a rock, you’ve probably heard a lot about AI lately, whether in general or as it relates to testing psychology. Maybe a colleague mentioned it at a conference. Maybe you’ve seen it come up in online groups or in professional listservs. And, if you’re being honest, you’re not entirely sure what to make of it.

You’re not alone. In our conversations with hundreds of clinicians, we’ve found that most neuropsychologists fall into one of three camps: those that are all-in and using AI daily, those who are cautiously curious about AI, and those who’ve already decided it’s not for them. For those that haven’t made the plunge just yet, folks tend to share some of the same set of concerns.

This article is about highlighting some of the most common misunderstandings (and maybe creating some new ones) so you can make an informed decision about whether, and how, AI might fit into your workflow.

Misconception #1: “AI is going to replace me.”

This is the big one, and it’s worth addressing head-on. The fear that AI will make neuropsychologists obsolete misunderstands what AI actually does well—and what it doesn’t.

AI excels at pattern recognition, processing large volumes of information quickly, and generating drafts of structured text. What it cannot do is build rapport, exercise clinical judgment in ambiguous situations, integrate decades of training and experience into a nuanced diagnostic impression, or navigate the deeply human aspects of an evaluation.

With every technological wave, there have been fears of replacement. Whether it was bank tellers and the ATM, or mathematicians and the calculator, or even so far back as teachers and the phonograph, technology was predicted to replace, rather than augment. (Today, there are more bank tellers than there were when ATMs came out.) These technologies replaced certain tasks, but allowed workers to focus on higher-order problems. AI has the potential to do the same for clinicians—handling the time-consuming, repetitive parts of your workflow so you can spend more time on what actually requires your expertise.

The bottom line: AI is a tool, not a replacement. The clinicians who will thrive are the ones who learn how to use it effectively, not the ones who ignore it entirely.

Misconception #2: “AI makes things up, so I can’t trust it.”

You’ve probably heard the term “hallucination” in the context of AI. It refers to instances where an AI generates information that sounds plausible but is factually incorrect. And yes, this is a real phenomenon. But the conversation usually stops there, which is a problem. It leads people to dismiss AI entirely rather than understanding the nuance of hallucinations, and misconception #1.

Here’s what’s worth knowing: hallucinations happen because of how AI models work. They don’t “know” things the way you do. They predict the most likely next word in a sequence based on patterns in their training data. Sometimes that prediction is wrong, especially when the model is asked about something outside its training or pushed to be overly specific. Hence, a general-purpose chatbot answering medical questions is very different from a purpose-built tool designed for a specific workflow, with guardrails, validation layers, and domain-specific training. And while the latter is much more likely to provide meaningful medical information, it's still not immune to producing errant diagnoses or analyses.

The bottom line: Hallucinations are a real challenge and yes, you should not trust AI at face value. But, as with any tool, there are right and wrong ways to use any tool—you wouldn’t try to hammer a nail using the handle, even if you could.

Misconception #3: “I’m not technical enough to use AI.”

There’s a perception that using AI requires coding skills, a computer science background, or at least a comfort level with technology that many clinicians feel they don’t have. One of the key aspects to AI is the fact that this was never true!

From the beginning AI tools were designed to be used in natural language—hence the “chat” in “ChatGPT.” You don’t write code—you write instructions in plain English. If you can describe what you need in a sentence or two, you can use most AI tools.

In fact, your clinical expertise is actually your biggest advantage. AI tools are only as good as the person guiding them. A neuropsychologist who understands what a well-written report looks like, what clinical nuance matters, and what language is appropriate for a given context will get far better results from AI than a tech-savvy person with no clinical background.

The bottom line: You don’t need to be technical. But… you do need to be willing to experiment. The learning curve is much shorter than you think.

Misconception #4: “Using AI is ethically questionable.”

This one makes sense. Neuropsychology is built on rigor, precision, and high ethical standards. So when AI enters the conversation, questions about patient privacy, data security, informed consent, and professional responsibility naturally follow. They should.

But it's worth distinguishing between two very different categories of AI tools. General-purpose models like ChatGPT are designed for broad consumer use — and yes, many of them do use input data to train future models. That's a legitimate concern, especially when patient information is involved. Tools built specifically for healthcare are a different story. They operate under strict compliance frameworks like HIPAA, with safeguards around data handling, storage, and access designed from the ground up—not bolted on after the fact.

Take a listen to a recent podcast from Dr. Jeremy Sharp.  

There's also the question of how these tools get used in practice. AI-assisted doesn't mean AI-replaced. A report still needs a clinician's review, clinical judgment, and signature before it goes anywhere. The tool does the drafting; you do the thinking. That distinction matters — both for the quality of the work and for the ethical standards behind it.

The bottom line: Ethical scrutiny is exactly the right instinct. The key is directing it toward evaluating specific tools and workflows rather than treating AI as a category to avoid entirely.

Misconception #5: “AI isn’t ready yet. I’ll wait until it’s perfect.”

Waiting for AI to be “perfect” is like waiting for your EHR system to be perfect. It’s never going to happen, and in the meantime, you’re spending hours on tasks that could be done in minutes. As you probably tell to your own clients, “perfection is the enemy of progress.”

AI today is imperfect—but it’s also remarkably capable. The clinicians who are experimenting with it now are building skills and workflows that will compound over time. They’re learning what works, what doesn’t, and how to integrate AI into their practice in a way that makes sense for them.

Steve Ballmer, then-CEO of Microsoft, famously dismissed the iPhone when it launched in 2007, saying no one would want it. The early iPhone was genuinely limited—no app store, no copy-paste, mediocre battery life. But the people who adopted it early shaped how it evolved, and they were miles ahead by the time the rest of the world caught up.

The same dynamic is playing out with AI. You don’t need to go all-in. But an experimental mindset—a willingness to try something new, evaluate it honestly, and iterate—will serve you far better than waiting on the sidelines.

The bottom line: AI will never be perfect. But it’s already good enough to make a meaningful difference in your practice—if you’re willing to take it for a test drive.

What Now?

If any of these misconceptions resonated with you, that’s a good sign. It means you’re thinking critically about AI rather than accepting or rejecting it on reflex.

Over the coming weeks, we’ll be publishing deeper dives into each of these topics—how AI hallucinations actually work, what to look for when evaluating an AI tool’s security and compliance, practical tips for getting started, and more. These articles are written for clinicians, not engineers, and our goal is simple: to help you make informed decisions about AI on your own terms.

In the meantime, if you have questions about AI in neuropsychology—or if there’s a misconception we didn’t cover that you’d like us to address—we’d love to hear from you.