Artificial Intelligence

AI is fast.
AI is fluent.
AI is confident.

But is it clinically trustworthy?

🤖 AI can generate answers in seconds.
But good healthcare decisions require more than speed.

As AI tools become more accessible across healthcare, the line between assistance and authority is starting to blur.

Don and Steve put artificial intelligence under the microscope—cutting through hype to examine how AI is actually performing in healthcare today.

They tackle the questions healthcare professionals are already facing:

🧠 Which clinical tasks can AI reliably support—and which still require human judgment?
 ⚠️ Is it actually helping clinicians — or quietly creating new risks?
 📋 When AI gives a medical answer, how often is it accurate… and how often is it confidently wrong?
 🎓 How should AI be used (or limited) when training the next generation of clinicians?

Then they put the evidence behind these claims to the test:

• AI will fully replace healthcare professionals
• Using AI improves patient safety
• AI generated medical answers are more accurate than humans
• Healthcare students should use AI

🏷️ Our listeners can get 10% off a new or upgraded subscription with code rvt1026 at checkout. 

******

TRC Healthcare Editor Hosts: 

• Stephen Small, PharmD, BCPS, BCPPS, BCCCP, CNSC 
• Don Weinberger, PharmD, PMSP

Guests:

• Vickie Danaher, PharmD
• John Turtle, PharmD, MBA

******

CE Information:

None of the speakers have anything to disclose. 

TRC Healthcare offers CE credit for this podcast for subscribers at our platinum level or higher. Log in to your Pharmacist’s Letter, Pharmacy Technician’s Letter,or Prescriber Insights account and look for the title of this podcast in the list of available CE courses.

Claim Credit

******

Clinical resources:

• FAQ: Artificial Intelligence in Pharmacy Practice
• Checklist: Responding to Med Errors
• Toolbox: Resources for Discussing Medical Misinformation
• Article: Weave AI Into Your Plan for Precepting Learners
• Toolbox: Preceptor’s Guide

******

Email us: rumorvstruth@trchealthcare.com

The content of this podcast is not intended to be a substitute for professional medical advice, diagnosis, or treatment.

Find the show on YouTube by searching for ‘TRC Healthcare’ or clicking here.

Learn more about our product offerings at trchealthcare.com.

Transcript

This transcript is automatically generated. 

00:00:05 Narrator

Welcome to Rumor vs. Truth, your trusted source for facts, where we dissect the evidence behind risky rumors and reveal clinical truths.

00:00:13 Narrator

Today, we'll see if claims around AI deserve a hard reset.

00:00:21 Steve Small

Now, Don, I've got a serious question for you here, and you can be honest.

00:00:24 Steve Small

Would you prefer to have an AI-generated co-host for the show instead of me?

00:00:30 Don Weinberger

It probably would be more efficient, but at the same time, it would flag my puns as low clinical value.

00:00:37 Don Weinberger

So that is a hard no.

00:00:40 Don Weinberger

You forever, Steve.

00:00:41 Steve Small

Thanks, Don.

00:00:41 Steve Small

And we never promised our humor is evidence-based, so—

00:00:45 Steve Small

And bad jokes might actually mean job security here.

00:00:48 Don Weinberger

Yep, exactly.

00:00:49 Don Weinberger

And rest assured, we are your fully human, non-AI generated hosts with the bad puns, everything.

00:00:55 Don Weinberger

With that, I'm Don the pharmacist.

00:00:57 Steve Small

And I'm Steve the pharmacist.

00:00:59 Don Weinberger

In this episode, we're looking at artificial intelligence and big claims around using large language models in healthcare to see what's real, what's exaggerated, and what actually matters in clinical practice.

00:01:10 Steve Small

Right, and before we plug into these claims, it's important to remind folks that this podcast offers continuing education credit for pharmacists, pharmacy technicians, prescribers, and nurses.

00:01:20 Don Weinberger

Just log into your Pharmacist’s Letter, Pharmacy Technician’s Letter, or Prescriber Insights account.

00:01:24 Don Weinberger

Look for the title of this podcast in the list of available CE courses.

00:01:28 Steve Small

And for the purposes of disclosure for today's podcast, none of the speakers have anything to disclose.

00:01:34 Don Weinberger

So, you know, Steve, glad we're doing this topic.

00:01:37 Don Weinberger

You know, artificial intelligence has met a lot of praise, and with that, lots of critique.

00:01:43 Don Weinberger

And it's pretty much ever since ChatGPT came out several years ago.

00:01:47 Don Weinberger

And even one of my family members calls AI automated incompetence.

00:01:52 Don Weinberger

So—

00:01:54 Don Weinberger

Is my aunt really fair with that?

00:01:57 Steve Small

Yeah, sounds kind of harsh, but while we debate AI's role in society, we know AI is already showing up in healthcare.

00:02:04 Steve Small

Things like clinical decision support, documentation, patient education, even drug info.

00:02:09 Steve Small

Also, AI isn't just chatbots here.

00:02:12 Steve Small

A lot of real-world examples include pattern recognition tools you might see in things like imaging or ECG screening, things like that.

00:02:21 Don Weinberger

And that's led to some growing assumptions about safety, accuracy, bias, and that famous question we all have: is AI coming for our jobs?

00:02:30 Steve Small

Yeah, those are some hair-raising concerns there.

00:02:33 Steve Small

And speaking of which, we'll also answer a listener's question about minoxidil for hair growth from our last episode.

00:02:39 Don Weinberger

Hair puns from that episode.

00:02:41 Don Weinberger

That was a great time, wasn't it?

00:02:43 Don Weinberger

All right, so let's go ahead and get to our first claim here.

00:02:46 Don Weinberger

And that actually goes back to our banter from the beginning.

00:02:49 Don Weinberger

And the claim is—

00:02:50 Don Weinberger

AI will fully replace healthcare professionals.

00:02:53 Steve Small

Yeah, this one sparks anxiety fast, especially in pharmacy, because when people see AI writing notes, checking interactions, or answering drug questions, it feels like replacement is right around the corner.

00:03:06 Don Weinberger

Right.

00:03:08 Don Weinberger

But across all the healthcare literature I could find, and especially in pharmacy, AI consistently shows up as a support tool and not a substitute.

00:03:16 Don Weinberger

Even the American Society of Health-System Pharmacists, ASHP,—

00:03:20 Don Weinberger

—statement on AI from 2025 says a common feature of current and future use cases is that they, AI, are designed to augment clinical pharmacy services, not replace the pharmacy workforce.

00:03:33 Steve Small

Yeah, and keep in mind, folks, while we're using a pharmacy example here because that's our wheelhouse, right, these same concerns show up for prescribers and nurses too.

00:03:41 Steve Small

And even though this sounds like a relief, are there good study examples backing up ASHP's claim here, Don?

00:03:48 Don Weinberger

Yeah, so—

00:03:49 Don Weinberger

If we look at pharmacy, for example, we do have studies comparing AI tools, such as large language models, to pharmacists.

00:03:57 Don Weinberger

And they show that AI does have strengths, but also weaknesses, like humans, right?

00:04:02 Don Weinberger

At least for right now.

00:04:03 Don Weinberger

And that pattern—strong in some tasks, weaker in others—is something we see across healthcare roles, not just pharmacists.

00:04:10 Don Weinberger

So to the studies.

00:04:12 Don Weinberger

One study I found compared and graded ChatGPT and clinical pharmacist answers to real-life medical cases—

00:04:18 Don Weinberger

—and standard test questions on a scale of 1 to 10, with 10 being correct.

00:04:22 Steve Small

Oh, pharmacist versus AI here, quite the showdown.

00:04:26 Steve Small

That should be good.

00:04:27 Don Weinberger

Right.

00:04:28 Don Weinberger

Well, ChatGPT performed equally well to the pharmacists regarding drug counseling, but it was weaker with prescription review, patient drug education, and recognizing and assessing adverse drug reactions.

00:04:41 Don Weinberger

And for perspective, ChatGPT scored in the fours to sixes on these categories while pharmacists were nine or above.

00:04:49 Don Weinberger

AI tended to miss issues with complex prescriptions or gave verbose answers to patient questions, which I think does make sense.

00:04:57 Steve Small

Yeah, and well done for those pharmacists based on those scores.

00:04:59 Steve Small

That's pretty good.

00:05:01 Don Weinberger

Yep.

00:05:02 Don Weinberger

Scoring one for the humans, right?

00:05:03 Don Weinberger

But keep in mind, the study was done in 2023, and the grading was subjectively done by five pharmacists.

00:05:10 Don Weinberger

And the study didn't mention any blinding.

00:05:13 Steve Small

Yeah, those are good caveats there.

00:05:15 Steve Small

And 2023, that can feel like a lifetime ago when it comes to this stuff.

00:05:20 Steve Small

And it can be hard to be up to speed with these studies when technology is evolving just so darn fast, right?

00:05:26 Don Weinberger

It seems like just exponential growth, right?

00:05:28 Don Weinberger

But on the flip side, evidence does seem to show it can help with other tasks and workflows.

00:05:33 Don Weinberger

So another study from 2023 looked at 30 pharmacists verifying 200 mock medication orders, half with AI's help and half without.

00:05:42 Don Weinberger

And some of the mock orders intentionally had errors they needed to spot.

00:05:46 Don Weinberger

So pharmacists made the correct decision 91% of the time without AI, but about 93 to 94% with AI.

00:05:55 Don Weinberger

Rates were especially good with uncertainty-aware AI, or AI that shares its confidence in its answer.

00:06:02 Don Weinberger

AI without this feature actually led to longer verification times compared to not using AI at all.

00:06:08 Don Weinberger

So it could slow things down.

00:06:10 Steve Small

So what I'm getting from that is it can help with certain tasks, especially more structured or rule-based ones.

00:06:16 Don Weinberger

Yep, exactly.

00:06:17 Don Weinberger

And for some pharmacy technicians in particular, that can raise real questions about how AI fits into verification, workflow, even automation.

00:06:27 Don Weinberger

Again, pointing to why thoughtful implementation actually matters.

00:06:30 Steve Small

I'm also so glad you mentioned the confidence factor here earlier.

00:06:33 Steve Small

AI can tell us what we want to hear and be confidently wrong about it, which could make you second guess things and, like you said, slow things down.

00:06:40 Steve Small

So it's a really good point.

00:06:41 Don Weinberger

Yeah.

00:06:42 Don Weinberger

Sounds just like my wife.

00:06:43 Don Weinberger

So—and this shows how we need to think about how AI is implemented.

00:06:49 Don Weinberger

If AI is implemented poorly without testing, AI can slow us down and make us less efficient.

00:06:54 Don Weinberger

Make sense?

00:06:56 Don Weinberger

Yeah.

00:06:57 Don Weinberger

All right, so let's go ahead and go back to that claim, which is AI will fully replace healthcare professionals.

00:07:03 Don Weinberger

And the verdict is—

00:07:09 Don Weinberger

This is a rumor.

00:07:11 Steve Small

Yeah, I would agree with ASHP here that current evidence shows AI will change how we do tasks, but not necessarily replace us.

00:07:18 Steve Small

And that applies broadly across healthcare teams, not just pharmacy staff here.

00:07:22 Steve Small

Okay.

00:07:23 Don Weinberger

Now, something fun, Steve.

00:07:25 Don Weinberger

I did pose this claim to Claude, which is our own AI software.

00:07:30 Don Weinberger

To see what it thinks the verdict should actually be.

00:07:32 Steve Small

Oh dear, I'm worried, Don.

00:07:34 Steve Small

I think it'll say true, but that's because I'm a total pessimist here.

00:07:39 Don Weinberger

Yeah, that's pretty aware working with you this long.

00:07:42 Don Weinberger

But actually, Claude called this rumor with conditions.

00:07:46 Don Weinberger

So pretty close.

00:07:49 Don Weinberger

Why?

00:07:49 Don Weinberger

Now, why did it say conditions?

00:07:50 Don Weinberger

So here's a piece of its explanation.

00:07:52 Don Weinberger

The claim—AI will fully replace healthcare professionals—is mostly a rumor, but it has a kernel of conditional truth.

00:08:01 Don Weinberger

So AI will replace or heavily shrink some tasks and some roles in some settings, especially the parts that are repetitive, rules-based, documentation-heavy, or primarily pattern recognition with clear guardrails.

00:08:15 Steve Small

Okay.

00:08:16 Steve Small

I could see its point, but I'd argue a lot of our roles aren't rules-based.

00:08:21 Steve Small

Healthcare is not black and white.

00:08:24 Steve Small

So we can't be fully replaced, is my thinking.

00:08:26 Steve Small

That said, AI won't replace healthcare professionals, but maybe those who use AI might replace those who don't.

00:08:35 Don Weinberger

Right.

00:08:37 Don Weinberger

Later on, Claude even said regulation and safety engineering push toward AI plus clinician, not AI instead of clinician.

00:08:47 Steve Small

Yeah, that is what makes me stand by our rumor verdict.

00:08:51 Steve Small

We need to keep the healthcare experts in the loop.

00:08:54 Steve Small

Even AI here doesn't think it should replace us, which might be the most reassuring thing we've heard all episode so far.

00:09:01 Don Weinberger

So as an example, we do use AI in our Rumor vs Truth editorial process, mostly as a review and refinement tool.

00:09:09 Don Weinberger

It can help us streamline ideas or even point out how bad our puns are.

00:09:13 Don Weinberger

And it's plenty busy with that.

00:09:14 Don Weinberger

But it doesn't create the content or make clinical decisions.

00:09:18 Don Weinberger

That usually requires the human expert, the human touch.

00:09:21 Steve Small

Yeah, and that is similar to how we use AI clinically, or how we should.

00:09:24 Steve Small

It helps us move faster on certain tasks, but it doesn't replace judgment.

00:09:29 Steve Small

We still rely on experts to decide what's accurate, relevant, and safe.

00:09:33 Steve Small

And when I think of clinical care too, the human experience is so nuanced and complicated, it's really hard to see how AI can fully replace things like empathy and compassion or navigate complex ethical issues we see every day—at least not yet.

00:09:49 Don Weinberger

Yeah, so I like that heartfelt point that you just gave, Steve.

00:09:53 Don Weinberger

But it's worth looking into ways AI can help you boost your efficiency in your daily tasks.

00:09:58 Don Weinberger

So we'll pitch our Artificial Intelligence in Pharmacy Practice FAQ on our website for a handy list of potential uses and considerations for AI in your work.

00:10:09 Don Weinberger

And one big theme we're seeing—and you can see it reflected in our AI resource—is that AI's first real foothold may be workflow: documentation, summarizing notes, helping teams spend less time clicking boxes—

00:10:24 Don Weinberger

—not replacing clinical judgment though.

00:10:25 Steve Small

Yeah, love that resource.

00:10:27 Steve Small

But whenever you introduce new tools into workflow, especially ones that touch patient care, the next question is always safety, right?

00:10:33 Steve Small

And it leads us to our next claim here, that using AI improves patient safety.

00:10:41 Don Weinberger

Right, so in fact—

00:10:43 Don Weinberger

Misuse of chatbots in healthcare is considered the top health technology hazard of 2026 by ECRI, which is the home of the Institute for Safe Medication Practices, or ISMP.

00:10:54 Steve Small

Yeah, so to help us untangle this one, I actually reached out to a colleague of mine, Dr. John Turtle, PharmD, MBA, who is an informatics pharmacist and the director of health system operations at VytlOne.

00:11:06 Steve Small

Let's see what he had to say on this.

00:11:11 Steve Small

Thanks for joining us today, John.

00:11:12 Steve Small

I'm so glad you could help us dive into this.

00:11:15 Steve Small

And I'd love to get your thoughts.

00:11:16 Steve Small

Do we know if AI improves any aspects of patient safety?

00:11:22 John Turtle

Hey, Steve, how you doing?

00:11:24 John Turtle

Good to talk to you.

00:11:27 John Turtle

You know, AI—the thing is, AI can detect safety risks earlier.

00:11:34 John Turtle

The first principle of, you know, all healthcare ethics is to do no harm.

00:11:39 John Turtle

But in and of itself, AI is not the risk.

00:11:42 John Turtle

It's how we use it, right?

00:11:44 John Turtle

So, as a clinician, it can help you identify what risks there may be in front of you, but it is ultimately your discretion.

00:11:55 John Turtle

And us as pharmacists, we serve as a critical redundancy in every single environment we practice in.

00:12:01 John Turtle

And so providers, nurses—

00:12:05 John Turtle

—you name it, are relying on our ability to verify whatever information is there and give a recommendation based on what's out in front of us.

00:12:12 John Turtle

So, you know, bolstering that use of AI can help patient safety, but ultimately it's not going to ever replace it.

00:12:25 Steve Small

Yeah, great point.

00:12:26 Steve Small

What can we do on our end to make that detection better?

00:12:30 Steve Small

I assume AI can't just do that automatically.

00:12:33 Steve Small

So what do we need to do on our end?

00:12:35 John Turtle

Well, so a lot of it is just making— you know, everything within clinical software applications is a workflow-based data input, right?

00:12:46 John Turtle

So making sure that your workflows are standardized, they reflect your policies and procedures, and that they are consistent.

00:12:55 John Turtle

So I always like to use the example of a vanco trough, right?

00:12:59 John Turtle

So you're dosing vanco, and this goes back to when I started as an intern—the quality in, quality out, right?

00:13:07 John Turtle

So if the information that you are using to perform your clinical assessment is bad, then your clinical assessment will be bad.

00:13:15 John Turtle

So the same applies for AI in that—

00:13:18 John Turtle

—whatever information is going into the system needs to be consistent and current in order for the AI to provide information back to you.

00:13:27 Steve Small

Great points there.

00:13:28 Steve Small

Yeah, I like that vanco example.

00:13:30 Steve Small

And are there any ways that AI can make safety worse that we know about, that we should be watching for?

00:13:36 John Turtle

Absolutely.

00:13:37 John Turtle

I mean, there's ways that if you're using an outdated data set—

00:13:42 John Turtle

—or any sort of outdated information—

00:13:45 John Turtle

—it’s going to learn AI.

00:13:47 John Turtle

There's three general modalities with AI.

00:13:50 John Turtle

There's neural networks, natural language processing, and predictive analytics.

00:13:54 John Turtle

All three of those—if you're using outdated information—steer clear.

00:14:02 Steve Small

Yeah, good points, good points.

00:14:04 Steve Small

And then what about alert fatigue too?

00:14:05 Steve Small

I've heard about concerns with that.

00:14:07 Steve Small

Does AI improve that?

00:14:09 Steve Small

Does it make it worse there too?

00:14:10 Steve Small

Or what's your take?

00:14:12 John Turtle

I mean, I think that I'm—Stephen—I'm bullish on AI.

00:14:16 John Turtle

I mean, I think that the more as pharmacists we can lean into it, the stronger that we can become, the more that we will be able to accomplish within our clinical work.

00:14:27 John Turtle

But I will say as far as clinical decision support—what I usually call alerts—right?

00:14:33 John Turtle

That's nothing new.

00:14:34 John Turtle

We've seen that.

00:14:35 John Turtle

I mean, I've seen clinical alerts since 2007 when I started in pharmacy.

00:14:41 John Turtle

And so—and you could think even back to before then.

00:14:44 John Turtle

I mean, Stephen, AI—we had spell check, right?

00:14:48 John Turtle

I mean, there was a computer telling you how to spell a word that it thought you were trying to spell a long time ago.

00:14:54 John Turtle

So as far as alert fatigue, there are always going to be things that pop up and tell the clinician, hey, watch out for this.

00:15:04 John Turtle

Now, I think that AI can—

00:15:07 John Turtle

—be more sensitive to the situation based on contextual workflows and contextual situations.

00:15:15 John Turtle

But at the end of the day, if the clinician is not paying attention to the alert, it is always a clinical discretion scenario.

00:15:24 Steve Small

Great points there.

00:15:25 Steve Small

Well, I think we got a lot of great material here.

00:15:28 Steve Small

I'm going to take this back to Don and see what he thinks.

00:15:30 Steve Small

But thank you, John.

00:15:31 Steve Small

This is awesome.

00:15:32 John Turtle

One more thing.

00:15:34 John Turtle

So the way I think of AI—if I'm a big Star Wars nerd, right?

00:15:39 John Turtle

AI is like—AI is like C‑3PO, right?

00:15:43 John Turtle

He's this—you know what?

00:15:45 John Turtle

C‑3PO is in a lot of those movies and he's always helping the good side over the bad.

00:15:51 John Turtle

But at the end of the day, you do not want C‑3PO flying the Millennium Falcon.

00:15:55 Steve Small

That's a good point.

00:15:58 Steve Small

I like that analogy there.

00:16:01 Steve Small

We'll leave it at that.

00:16:02 Steve Small

And I'll bring that back to Don.

00:16:04 Steve Small

I'm sure he loves the Star Wars reference too.

00:16:05 Steve Small

So we'll see.

00:16:06 Steve Small

Thank you, John.

00:16:08 John Turtle

Have a good day.

00:16:13 Don Weinberger

Wow, that really was a great interview—and always appreciate a good Star Wars reference.

00:16:19 Don Weinberger

So I knew you would for that.

00:16:21 Don Weinberger

Yeah.

00:16:22 Don Weinberger

So we really struggle to find, you know, robust randomized data that show that AI improves or harms patient outcomes.

00:16:31 Don Weinberger

But we have growing observational evidence that show it can go either way, right?

00:16:34 Don Weinberger

It's important to consider that AI may catch a dosing error at scale, miss the one detail that a human would catch in seconds.

00:16:43 Steve Small

Right, and we talked about some of AI's weaknesses earlier that could lead to safety issues.

00:16:47 Steve Small

But on the other hand, I also found some systematic reviews that showed AI can be helpful in other ways with safety, such as detecting possible errors after they've occurred—

00:16:57 Steve Small

—or processing large amounts of error report data to maybe find patterns, provide solutions, things like that.

00:17:03 Steve Small

So it really depends on the situation here and how it's used, kind of like what Dr. Turtle was talking about.

00:17:09 Don Weinberger

Another safety use case I thought was interesting is how AI may be used in automation to help limit pharmacist exposure to hazardous meds, which definitely would be a positive for safety.

00:17:20 Steve Small

Yeah, that's a unique use case and the list could go on and on.

00:17:23 Steve Small

So when it comes to this claim—

00:17:25 Steve Small

—that using AI improves patient safety—the verdict is evidence is mixed.

00:17:33 Steve Small

And what's important to realize with any technology, right, is that it depends on how you use it.

00:17:38 Steve Small

Whenever I have talked to students and residents about healthcare technology, I tell them to treat it like a power tool.

00:17:45 Steve Small

You can use an electric saw, for example, to efficiently build something amazing, or you can accidentally cut off all your fingers.

00:17:52 Steve Small

It depends on how you use it.

00:17:54 Don Weinberger

Right.

00:17:55 Don Weinberger

And you could be building something out of fingers, though—then it'd be working, right?

00:17:57 Don Weinberger

But that is definitely a memorable and grisly way to think about it, Steve.

00:18:02 Steve Small

Didn't mean to jump scare you there, Don.

00:18:04 Steve Small

But it's the same with AI.

00:18:06 Steve Small

It can help us be more efficient, but it can also hurt people if we use it improperly or without safeguards.

00:18:13 Don Weinberger

Right.

00:18:13 Don Weinberger

I agree.

00:18:13 Don Weinberger

And staff should always check their employer's policies on approved software and proper use before using an AI program—

00:18:20 Don Weinberger

—in their practice.

00:18:21 Steve Small

Not all programs are alike.

00:18:23 Steve Small

And prior testing is really important to make sure AI products are up to the task.

00:18:27 Steve Small

And the practical start, perhaps, is using AI to start with low‑risk—

00:18:32 Steve Small

—structured tasks first, like we talked about earlier, so you can build your trust in it.

00:18:36 Steve Small

And then if it does well, only then do you move into the high‑stakes clinical decisions.

00:18:42 Don Weinberger

And continue keeping patient safety top of mind with all technology.

00:18:45 Don Weinberger

So you can use our Responding to Med Errors checklist if an event occurs to make sure you gather all the facts, address the error, and reduce the risk in the future.

00:18:54 Steve Small

Yeah, and take a look at the show notes or description.

00:18:56 Steve Small

We've linked directly to that resource in Pharmacist’s Letter, Pharmacy Technician’s Letter, and Prescriber Insights, as well as our Artificial Intelligence in Pharmacy Practice FAQ we mentioned earlier.

00:19:07 Don Weinberger

Right.

00:19:07 Don Weinberger

If you're a new subscriber, don't miss out on these resources.

00:19:09 Don Weinberger

Sign up today to stay ahead with trusted insights and tools.

00:19:15 Don Weinberger

Now, I'm sure we know at least one person who has asked AI a medical question.

00:19:20 Don Weinberger

Perhaps we've done it ourselves.

00:19:23 Don Weinberger

So that leads to the next important claim, which is AI‑generated medical answers are more accurate than humans.

00:19:30 Don Weinberger

Now, when you look at this in two parts—questions patients ask AI versus questions healthcare workers pose to AI.

00:19:38 Steve Small

I agree.

00:19:39 Steve Small

Two different beasts that we're talking about.

00:19:41 Steve Small

And I'm sure many of us have seen how AI programs will add a disclaimer, for example, saying this is for informational purposes only—

00:19:49 Steve Small

—for medical advice or diagnosis, consult a professional.

00:19:53 Steve Small

So knowing that, are there any good study examples looking at this with patients?

00:19:57 Don Weinberger

Yes.

00:19:58 Don Weinberger

So keep in mind the studies out there on this are hard to interpret overall.

00:20:02 Don Weinberger

They use different AI programs, methods, and can involve very specific settings.

00:20:08 Don Weinberger

For example, like patient questions on a specific surgery procedure.

00:20:12 Steve Small

Oh yeah, that is indeed specific.

00:20:14 Steve Small

A really good point.

00:20:16 Don Weinberger

Right.

00:20:16 Don Weinberger

So—but one example—

00:20:19 Don Weinberger

—I saw was a randomized UK study in early 2026 looking at patients using AI versus a traditional internet search.

00:20:27 Don Weinberger

The search would be the control here to identify health condition and course of action.

00:20:32 Don Weinberger

The researchers gave around 1,300 participants various medical scenarios vetted by physicians, which they carried out with AI versus control, compared to what the physicians would have done.

00:20:45 Steve Small

Yeah, that's a pretty interesting design.

00:20:47 Steve Small

What did they find out from that?

00:20:48 Don Weinberger

When tested by researchers themselves, AI correctly identified conditions in 95% of cases and proper courses of action in 50%.

00:20:57 Don Weinberger

On the other hand, this dropped to 35% for conditions and 44% for courses of action when the patients actually used AI, which was about the same as a traditional internet search.

00:21:10 Don Weinberger

So a lot of it depended on what the patient told the program and the risk of leaving out key information.

00:21:16 Steve Small

So in a perfect setting with researchers, it does relatively well, but with patients, maybe not so much.

00:21:20 Steve Small

That can make sense.

00:21:22 Steve Small

So it has potential, but it likely depends on the user.

00:21:26 Don Weinberger

Yeah, exactly.

00:21:27 Don Weinberger

And for healthcare professional questions, one large real‑world study looked at 300 actual questions that were previously answered by a drug information service.

00:21:37 Don Weinberger

When those same questions were given to ChatGPT and Google’s Gemini, only 19% of AI responses were fully accurate—

00:21:45 Don Weinberger

—and supported by reliable references when graded by a senior pharmacy intern and a group of drug information pharmacists.

00:21:52 Steve Small

Yeah, that begs the question, Don—how many of these answers were just flat‑out wrong then?

00:21:57 Don Weinberger

Well, you never might surprise you.

00:22:00 Don Weinberger

About 15% of ChatGPT's answers were completely inaccurate, having different conclusions than the drug information pharmacist and using references that didn't really support the answer.

00:22:14 Don Weinberger

And one example the authors gave was a drug information question asking: does Orencia vial, auto‑injector, and syringe formulations contain sodium phosphate?

00:22:25 Don Weinberger

The pharmacist did confirm that it has it, but AI incorrectly said the auto‑injector formulation did not have sodium phosphate.

00:22:33 Steve Small

Oh, that is quite off.

00:22:35 Don Weinberger

Yes, yeah.

00:22:36 Don Weinberger

And most of the rest were partially correct or incomplete, which was seen in other studies too.

00:22:44 Don Weinberger

I do have a personal example from when we were creating our next CE presentation, and one of it has to do with insulin.

00:22:52 Don Weinberger

And I asked AI to help me design a plan for a patient who's switching from a higher concentration of insulin to a lower concentration of insulin.

00:23:01 Don Weinberger

So it gave a good answer as far as how to convert it, and it gave good references like the American Diabetes Association and other primary references that I was able to chase to the real answer.

00:23:13 Don Weinberger

Now, where it kind of messed up a little is it recommended an insulin product that doesn't exist anymore.

00:23:22 Don Weinberger

So we kind of—yeah.

00:23:24 Don Weinberger

So if you were to run to the prescriber or recommend to the patient, you may not be looking too informative if you recommend something that doesn't exist.

00:23:31 Don Weinberger

So that's just my personal example I experienced.

00:23:34 Steve Small

Yeah, that's a good one.

00:23:35 Don Weinberger

Yeah.

00:23:35 Don Weinberger

All right, so let's go ahead and circle back to that claim, which is AI‑generated medical answers are more accurate than humans.

00:23:41 Don Weinberger

And the verdict is—

00:23:48 Don Weinberger

Rumor with conditions.

00:23:50 Steve Small

Yeah, so AI is very good at retrieving and summarizing information, but it's weaker at weighing competing risks, recognizing when information is missing, et cetera.

00:23:59 Steve Small

So that was a good example, Don.

00:24:01 Steve Small

Maybe you didn't know that a certain product was off the market.

00:24:04 Steve Small

So what was AI's verdict to this one?

00:24:07 Steve Small

Did you plug this in?

00:24:08 Don Weinberger

I—you know—I did.

00:24:10 Don Weinberger

So when plugging this question into Claude again, it said evidence is mixed.

00:24:14 Don Weinberger

So we—it’s—

00:24:16 Don Weinberger

—still a loft.

00:24:17 Don Weinberger

Quoting it here, it said AI‑generated medical answers are sometimes more accurate than some humans in some settings, but not reliably more accurate than humans overall, and often not better than experts.

00:24:31 Don Weinberger

So a little humble pie there.

00:24:33 Steve Small

Yeah, based on that explanation, I think we're safe sticking to our own verdict on this one.

00:24:37 Don Weinberger

Right.

00:24:37 Don Weinberger

So when I told it that “evidence is mixed” isn't an option, Claude leaned toward rumor with conditions instead.

00:24:45 Don Weinberger

So we're roughly on the same page.

00:24:47 Steve Small

Yeah, so it kind of goes back to those prompts and what you tell it.

00:24:50 Steve Small

They can really affect the outcome there.

00:24:53 Don Weinberger

And we know AI programs may not just hallucinate and give wrong answers—

00:24:58 Don Weinberger

—they can be confidently wrong too.

00:25:00 Don Weinberger

So it's a good idea to prioritize using AI programs that tell the user the program's confidence in an answer, or simply ask AI, how confident are you in that answer?

00:25:11 Steve Small

Kind of putting AI on the spot there.

00:25:12 Steve Small

I like it.

00:25:13 Steve Small

And it's key to keep in mind that AI can be prone to bias depending on the sources it uses.

00:25:20 Steve Small

Yale School of Medicine had an article in 2024 on this entitled Bias In, Bias Out, which is a good way to sum it up.

00:25:28 Steve Small

So if it's using flawed data or misinformation to generate an answer, the concern is it can lead users down the wrong path.

00:25:35 Steve Small

And we have to keep that in mind.

00:25:37 Don Weinberger

And I'm going to pitch another resource of ours because we have quite a few on this.

00:25:41 Don Weinberger

So don't let AI send you or your patients down the rabbit hole, really.

00:25:44 Don Weinberger

You can see our resource called Resources for Discussing Medical Misinformation, a chart to help navigate any questionable information that may come from AI or other sources.

00:25:54 Steve Small

Yeah, that's a great resource, Don.

00:25:56 Steve Small

But with these benefits and risks, where does AI fall when it comes to educating the next generation of healthcare professionals?

00:26:04 Steve Small

Our next claim, in fact, is healthcare students should use AI.

00:26:08 Steve Small

And to get a grasp on whether or not AI belongs in healthcare education, I posed this question to our fellow editor, Vickie Danaher, PharmD.

00:26:16 Steve Small

She's written a lot of our great content on AI, including an article in Pharmacist’s Letter on this very topic.

00:26:22 Steve Small

So let's see what she had to say.

00:26:27 Steve Small

Thank you for joining us, Vickie.

00:26:28 Steve Small

And I'm curious—what is your take on this claim that healthcare students should use AI?

00:26:33 Steve Small

What benefits do you see, and are there risks?

00:26:36 Vickie Danaher

Yeah, definitely.

00:26:37 Vickie Danaher

Thanks for having me, Steve.

00:26:38 Vickie Danaher

So AI is already becoming a part of healthcare, as you guys have talked about already.

00:26:44 Vickie Danaher

So ignoring it or not talking about it with students or using it in learning experiences isn't really going to prepare them for the real world.

00:26:53 Vickie Danaher

We know that patients are using AI—they're looking for it for questions on drug information, health information, and they're coming to us to ask about it.

00:27:03 Vickie Danaher

So we want to make sure that students are prepared for the real world that they're going to be living in and working in.

00:27:11 Steve Small

Yeah, the cat's already out of the bag.

00:27:13 Vickie Danaher

Definitely.

00:27:14 Steve Small

So what are good examples you suggest listeners should think about if they want to use or integrate AI into teaching?

00:27:22 Vickie Danaher

So as a pharmacist, one of the main examples that comes to mind for me is drug information questions.

00:27:29 Vickie Danaher

And so patients are coming to us with questions, or clinicians have questions.

00:27:33 Vickie Danaher

And there's really kind of two ways that we can work with our students to approach those questions using AI.

00:27:40 Vickie Danaher

So one way might be for students to prepare a response to a drug information question in the traditional way—

00:27:46 Vickie Danaher

—consulting primary literature, consulting databases, writing up the response on their own.

00:27:52 Vickie Danaher

And then using AI to generate its own response and comparing those two versions.

00:27:58 Vickie Danaher

So was there something that AI missed, or was there something the student missed, or thought about differently?

00:28:04 Vickie Danaher

And that can really be an area to improve their clinical thinking or clinical judgment.

00:28:11 Vickie Danaher

Another option would be to have the student use AI to start off the response to the drug information—

00:28:17 Vickie Danaher

—using it as a starting point and critiquing what it comes up with up front, then revising and building upon it to create a solid answer.

00:28:29 Vickie Danaher

But I think those are two different ways, but the outcome is the same.

00:28:34 Vickie Danaher

You're still looking at the response or information AI is producing.

00:28:40 Vickie Danaher

You're evaluating it, critiquing it, to ensure the answer is solid.

00:28:45 Steve Small

Lots of options—and both lead to great learning.

00:28:47 Steve Small

So what would you say or recommend to listeners who may still be hesitant to use AI with learners?

00:28:54 Vickie Danaher

Yeah, I totally understand the hesitation, and there are many issues surrounding AI.

00:29:01 Vickie Danaher

But I would say the more that I've used it and experimented with it, you build that understanding and awareness of its capabilities and its limitations.

00:29:12 Vickie Danaher

You as well as students will learn what needs to be verified.

00:29:17 Vickie Danaher

And you know that you can't trust it completely, but you can use it as a tool to support the work you're doing.

00:29:24 Steve Small

Excellent, excellent thoughts.

00:29:26 Steve Small

And I'll be looking forward to sharing these with Don and hopefully calming some of his fears around AI.

00:29:30 Steve Small

We'll see.

00:29:31 Steve Small

But thank you for joining us today.

00:29:32 Vickie Danaher

Sounds good.

00:29:36 Don Weinberger

Wow—another great interview, Steve.

00:29:38 Don Weinberger

But I'm glad to actually have Vickie on our podcast.

00:29:42 Don Weinberger

It's great to see her.

00:29:44 Don Weinberger

But I have to circle back to what you capped on at the end here.

00:29:47 Don Weinberger

And who said I was afraid of AI?

00:29:49 Don Weinberger

You know—speak for yourself.

00:29:53 Steve Small

Oops, I may have spoken too soon.

00:29:54 Steve Small

Don, I put words in your mouth.

00:29:56 Don Weinberger

Right.

00:29:56 Steve Small

But okay.

00:29:56 Steve Small

We can say here and now—we can document for the record—it is a rumor that Don is afraid of AI.

00:30:04 Don Weinberger

Right, and this is recorded, so it's officially on the record, right?

00:30:07 Don Weinberger

And you may have noticed we brought in two experts this time—voices for our podcast.

00:30:12 Don Weinberger

It was actually new for us.

00:30:13 Steve Small

Yeah, and that's intentional.

00:30:15 Steve Small

With AI, the best outcomes happen when you keep multiple experts in the loop.

00:30:19 Steve Small

So we figured we should practice what we preach.

00:30:22 Don Weinberger

And honestly, that perspective really showed here.

00:30:25 Don Weinberger

I really liked her suggestions for integrating AI into education.

00:30:28 Don Weinberger

The idea is that it also teaches us how to balance AI's benefits and risks in everyday use.

00:30:32 Steve Small

Exactly.

00:30:33 Steve Small

So when it comes to this claim that healthcare students should use AI, the verdict is true with conditions.

00:30:44 Don Weinberger

And I agree—AI is not going anywhere.

00:30:47 Don Weinberger

So we shouldn't turn a blind eye to it.

00:30:49 Don Weinberger

And evidence suggests students are using it.

00:30:53 Steve Small

Right, and we're in a prime position to guide learners on proper use—and knowing how and when to dig deeper when AI provides suggestions.

00:31:02 Steve Small

We don't want learners approaching AI with fear—maybe just a healthy amount of vigilance instead.

00:31:07 Don Weinberger

We don't want learners outsourcing their thinking.

00:31:10 Don Weinberger

We want them sharpening it, right?

00:31:11 Don Weinberger

So if you want more structure and examples, you can read Vickie’s excellent article in the January 2026 issue of Pharmacist’s Letter, Pharmacy Technician’s Letter, and Prescriber Insights to help with ideas for incorporating AI when teaching learners.

00:31:25 Steve Small

Yeah, and with that said, you don't need AI to tell you what to do if you're enjoying the show.

00:31:29 Steve Small

We've got the human‑verified answer right here.

00:31:33 Don Weinberger

Yep, exactly.

00:31:33 Don Weinberger

And are you a subscriber?

00:31:34 Don Weinberger

Don't forget to claim CE credit for this episode.

00:31:37 Steve Small

And not a subscriber yet or thinking about upgrading?

00:31:40 Steve Small

Access more trusted clinical insights and save 10% with our exclusive listener promo code RVT1026 at checkout.

00:31:50 Don Weinberger

Check out details and links in the show notes below.

00:31:52 Don Weinberger

Don't miss out.

00:31:53 Steve Small

And with that, the bottom line truth here today is that AI can be fast and fluent and confident, but it can still be clinically unsafe without proper oversight.

00:32:04 Don Weinberger

And AI is a powerful copilot.

00:32:07 Don Weinberger

Just don't put it on autopilot.

00:32:09 Steve Small

Exactly.

00:32:09 Steve Small

And as healthcare professionals, we have the knowledge and know‑how to assess AI suggestions to make sure its outputs are applied appropriately and safely to patient care.

00:32:19 Don Weinberger

Okay, so now it's time to stop promoting AI and instead see what our listeners prompted us from the Rumor vs Truth mailbag.

00:32:26 Don Weinberger

We have an audience question from our last episode about hair loss that came in through our “Send Us a Text” link in the podcast show notes.

00:32:35 Don Weinberger

And what they're asking is: is minoxidil effective for growing a better beard?

00:32:40 Steve Small

Oh, interesting question.

00:32:41 Steve Small

You and I don't have trouble with this.

00:32:43 Steve Small

Don—our beards are crazy—but this may not be a totally hair‑brained idea.

00:32:48 Steve Small

I can see where this is coming from.

00:32:50 Steve Small

Now, first, to be clear, topical minoxidil is only FDA‑approved for regrowing hair on the top of the scalp.

00:33:03 Steve Small

It even has a warning saying do not apply to other parts of the body.

00:33:10 Steve Small

So keep in mind here, folks, this hair growth idea is definitely an unapproved, off‑label use when it comes to the beard.

00:33:10 Steve Small

But that said, a randomized, placebo‑controlled study from 2016 looked at using 0.5 mL of 3% topical minoxidil lotion twice daily to the beard area in 46 men who wanted fuller facial hair.

00:33:25 Steve Small

And several physicians then graded how well their beards looked in photos at the end of the trial.

00:33:31 Don Weinberger

So 3% minoxidil lotion—okay, interesting.

00:33:34 Steve Small

How well did it do?

00:33:36 Steve Small

Yeah, the minoxidil patients did have better subjective beard scores compared to the placebo group, although the study didn't really say by how much.

00:33:43 Steve Small

It was kind of unclear.

00:33:45 Steve Small

And the minoxidil group also rated their own beards more highly by the 16‑week mark.

00:33:50 Steve Small

And they reported side effects that were mild.

00:33:53 Steve Small

But keep in mind, we don't carry minoxidil 3% lotion like you were hinting at, Don.

00:33:58 Steve Small

We don't carry that in the U.S.

00:33:59 Steve Small

And this study was small.

00:34:01 Steve Small

So it's kind of difficult to apply these results here.

00:34:04 Don Weinberger

Yeah, so thank you for specifying that.

00:34:06 Don Weinberger

So I wouldn't say you're just splitting hairs there, right?

00:34:11 Don Weinberger

Those are good things to point out.

00:34:12 Steve Small

Yeah, and you might actually get a kick out of this interesting 2024 case report involving identical twin males.

00:34:19 Steve Small

One twin used 5% topical minoxidil once daily on the beard for 16 months, while the other didn't.

00:34:26 Steve Small

So kind of an interesting control group.

00:34:29 Steve Small

The minoxidil twin did show greater subjective beard density and hair growth at that 16‑month mark—

00:34:34 Steve Small

—while reporting only mild local effects like dryness and some increased body hair.

00:34:41 Steve Small

And they also had some initial shedding, which you can expect with minoxidil.

00:34:45 Steve Small

But looking at the photos from this study, even though there was subjective improvement and maybe some improved hair density, it wasn't a lumberjack‑level beard by any means.

00:34:54 Steve Small

In fact, I'd say you and I have fuller beards just on this podcast.

00:34:59 Don Weinberger

Well, yeah.

00:34:59 Don Weinberger

Well, I use mine to hide my double chin.

00:35:01 Don Weinberger

So—but—

00:35:03 Don Weinberger

So is minoxidil the answer to get a big, bushy beard?

00:35:07 Steve Small

I would say that it can improve beard hair density for some people.

00:35:12 Steve Small

And we do have limited controlled trial data and case‑level data to support that.

00:35:17 Steve Small

But it's still off‑label.

00:35:18 Steve Small

And we should look at long‑term outcomes and side effects before using this routinely.

00:35:23 Steve Small

I would say if somebody is considering this, it's a good opportunity for healthcare professionals to set realistic expectations.

00:35:31 Steve Small

You can review the risks and help patients make an informed decision.

00:35:34 Don Weinberger

Okay, and thank you for that.

00:35:36 Don Weinberger

And this is just the kind of audience question we love.

00:35:39 Don Weinberger

It's practical, unexpected, and a little hairy to figure out.

00:35:44 Don Weinberger

Now, if you've got an AI‑related question from this episode—or a technology rumor you want us to fact‑check—send it our way.

00:35:51 Steve Small

Yeah, and we also use your suggestions to plan our episodes.

00:35:53 Steve Small

So email us at rumorvstruth@trchealthcare.com or send us a text right from the podcast show notes.

00:36:00 Don Weinberger

And before you go, claim CE credit and access evidence‑based resources from Pharmacist’s Letter, Pharmacy Technician’s Letter, or Prescriber Insights.

00:36:09 Steve Small

And if you're not yet a subscriber or want to upgrade, save 10% with our exclusive listener code RVT1026 at checkout.

00:36:17 Steve Small

There's an easy link in the show notes.

00:36:19 Don Weinberger

And already a subscriber?

00:36:21 Don Weinberger

Tap the “Claim Credit” link in the show notes or search your CE organizer for this episode.

00:36:28 Steve Small

And join us next time, where we'll separate smart prescribing from stubborn myths around antimicrobial stewardship.

00:36:34 Don Weinberger

You could say that topic is hard to resist—and that's kind of the problem, right?

00:36:40 Steve Small

Exactly.

00:36:40 Don Weinberger

So thanks for joining us on Rumor vs Truth, your trusted source for facts, where we dissect the evidence behind risky rumors and reveal clinical truths.

00:36:48 Don Weinberger

See you next time.