When You Ask ChatGPT If Your Memory Is Okay

There’s something almost logical about asking AI whether your thinking has changed. These tools can help write a legal brief, debug code, draft a salary negotiation email, plan an entire vacation itinerary, and settle a dinner table argument. Of course people would also wonder if AI could tell them whether the fogginess they’ve been feeling lately is something worth paying attention to.

It makes sense. And you’re far from alone in trying it. A West Health–Gallup survey conducted in late 2025 found that over 66 million Americans have used AI chatbots for health information or advice. It’s fast, it’s private, and it’s there at two in the morning when the worry won’t quiet down and you’re not sure what else to do with it.

But here is where it gets complicated. These same tools that can do so much confidently, quickly, and convincingly, cannot actually evaluate your cognitive health. Not even close.

AI can raise a concern. It cannot assess one.

I’m Dr. Rebecca Steele, a licensed clinical psychologist with advanced training in neuropsychology. I work with people in Garden City and across Long Island, and I see clients virtually throughout New York State and Georgia. What I want to offer here isn’t a warning to put your phone down and stop researching. It’s an honest look at what AI can and cannot do when it comes to your brain health, and what a real next step looks like when something doesn’t feel right.

Why So Many People Are Turning to AI for Brain Health Questions

Healthcare can feel hard to access, expensive, and slow. And brain health questions carry a particular weight, because sitting with the possibility that something might be worth looking into isn’t a comfortable place to be. Typing symptoms into a chatbot at midnight feels easier than scheduling an appointment somewhere and waiting weeks just to get five minutes with someone.

Here’s what I want you to know: a free 10-minute consultation with me takes less time than the rabbit hole you probably just fell into. You don’t have to wait three weeks, and you don’t have to walk in already knowing what’s wrong. That conversation is here when you’re ready.

What’s interesting is that most people aren’t turning to AI because they think it’s a doctor. According to the same Gallup study, they’re turning to it because they want answers quickly (71%), want more information (71%), or are simply curious what it would say (67%). Those are completely reasonable impulses. The problem isn’t the question. It’s when the response gets treated as a clinical answer, because those are not the same thing.

What AI Can Actually Do (And What It Cannot)

AI tools like ChatGPT or Gemini are trained on vast amounts of text. They can summarize information, explain concepts in plain language, and produce something that sounds reasonable and confident in response to almost anything you type. For a lot of everyday questions, that’s genuinely useful.

A 2024 study in the Journal of Alzheimer’s Disease Reports looked at ChatGPT as a diagnostic aid in Alzheimer’s disease. The researchers found that, when given detailed case information, it produced symptom summaries and diagnostic impressions that matched those of dementia specialists, which sounds promising. But they were explicit about what that means: ChatGPT should be seen as an adjunct to clinical judgment, not a stand‑alone diagnostic method. Even in a study designed to showcase a legitimate role for AI in the Alzheimer’s diagnostic process, the authors were clear that it cannot replace a clinician’s evaluation.

And here is a big part of why. By default, a chatbot has no access to:

• Your full medical and psychiatric history
• Your cognitive baseline, including how you functioned at your best before the changes you’re noticing now
• Your sleep quality, hormone and thyroid status, or how your medications interact with each other
• Your day‑to‑day stress, mood, and how anxiety or depression are showing up in your thinking, focus, and memory
• Any direct observation of how you actually perform across different cognitive tasks, including effort, behavior in the room, and how you respond to being challenged

Even if you choose to connect health data to an AI tool, it still can’t watch you think in real time or replace a structured, in‑person evaluation.

Research published in BMC Medical Education found that AI chatbots tend to answer decisively even in situations of genuine diagnostic uncertainty, often overestimating how confident they should be in their answers. In plain terms: a chatbot may sound very sure in exactly the kind of situation where a clinician would slow down, ask more questions, and gather more information before saying anything.

A concerning result from an online tool is not a diagnosis. But it is also not nothing. At best, it is a starting point for a real conversation with someone who can actually evaluate what is going on.

Cognitive Screening and Assessment

What Actually Happens During a Cognitive Screening

When you come to see me for a cognitive screening in New York, what happens is a different process entirely from typing symptoms into a search bar.

When I work with someone, I’m looking at the full picture of who you are, not just the symptoms you can articulate at midnight when you’re anxious. That includes:

• How you perform across multiple cognitive domains, including memory, attention, processing speed, executive functioning, and language
• Your personal history, including medical, psychiatric, sleep, and lifestyle factors
• Your baseline, because a meaningful change from your personal best looks very different from one person to the next
• Whether mood, anxiety, or stress might be contributing to what you’re experiencing, because they often are

This matters because many of the things that look like memory loss or cognitive decline could actually be something else entirely. Sleep deprivation, perimenopause, untreated depression, certain medications, thyroid imbalances, and chronic stress can all produce cognitive symptoms that a chatbot has no way of distinguishing from early neurological changes. A trained clinician can make that distinction.

If you’re in or around Garden City on Long Island, or anywhere in New York State or Georgia and curious about what a virtual session looks like, I’d love to hear from you. A free 10-minute consultation is a good place to start.

The Problem With Confident Responses to Uncertain Questions

AI tools are becoming less likely to remind you of their own limits. A Stanford-led study, reported in outlets like Rolling Stone and MIT Technology Review, found that in 2022 about 26 percent of chatbot responses to health questions included some kind of disclaimer that the tool was not a substitute for professional medical advice. By 2025, fewer than one percent of responses contained that reminder.

That shift matters. It means people are receiving health information from AI with fewer and fewer signals that they should treat it with any caution at all.

I’m not sharing this to make you distrust everything you read. I’m sharing it because cognitive health is one of those areas where the specifics matter enormously, and where a confident-sounding generalization can genuinely lead you in the wrong direction.

Forgetting where you left your keys is almost never what you fear it might be. But sometimes changes in memory, word-finding, or mental sharpness do deserve a real look. The way to know the difference isn’t to ask AI. It’s to sit down with a clinician who can actually evaluate you as an individual.

When an AI Chatbot Response Should Send You to a Clinician

If you’ve already gone down the AI rabbit hole on this, here is how I’d encourage you to think about what you found.

Treat the response as a signal, not a verdict. If something surfaced that made you pause, honor that pause, not because the chatbot gave you a diagnosis, but because your own noticing came first. AI was just the first place you took it.

It may be worth reaching out for a proper evaluation if you’re noticing any of the following:

• Memory changes that feel different from ordinary forgetfulness, particularly if people close to you have noticed something too
• Difficulty following conversations, losing track mid-sentence, or struggling to find words you know well
• Trouble with tasks that used to feel automatic, like managing finances, following a recipe, or navigating a familiar route
• Mental fatigue that feels out of proportion to your activity level or sleep
• A general sense that something has shifted, even if you can’t quite put your finger on what

None of these is an automatic sign of something serious. But they are the kinds of things that deserve a proper clinical look rather than a chatbot’s best guess based on a few lines of text.

In my practice, cognitive screening and assessment gives us a clear picture of where you actually are, in a way that accounts for your individual history and baseline. Sometimes that picture is genuinely reassuring. Sometimes it points toward something we can work on together. Either way, what you walk away with is information that is actually about you, not a statistical pattern generated from a prompt.

AI as a Starting Point, Not a Stopping Point

I want to be clear that I’m not dismissing what AI can offer. The Gallup research found that about 46 percent of people who used AI for health information said it made them feel more confident going into a conversation with a provider, and 22 percent said it helped them identify issues earlier. That is real value. Used well, AI can help you articulate what’s been bothering you, prepare questions for an appointment, and feel less alone with something that’s been weighing on you.

What it cannot do is replace the judgment that comes from actually sitting with you, reviewing your history, and watching how you perform across a range of cognitive tasks. Your brain is not a general case. It is your specific baseline, your specific history, your specific life, and it deserves to be evaluated as such.

If something you found through AI, or something you’ve simply been noticing on your own, has been sitting with you, I’d encourage you to take the next step. Cognitive screening in New York and Georgia is available through my practice both in person and virtually, and it doesn’t have to mean something is wrong.

Book a free 10-minute consultation here. I see clients in person at my Garden City office on Long Island, and virtually throughout New York State and Georgia.