Special Edition: Dr. AI Chatbot? What to Know Before You Ask artwork

Special Edition: Dr. AI Chatbot? What to Know Before You Ask

The NewsWorthy

August 1, 2026

More people are turning to AI chatbots with health questions, and they're getting quick, confident answers. But how much should you actually trust what AI tells you?
Speakers: Matthew Callstrom, Celine Gounder, Erica Mandy
**Matthew Callstrom** (0:00)
Anything that informs patients about their own ill, I think is really helpful.

**Celine Gounder** (0:05)
In one of the cases, the management of the patient it was suggesting would, in real life, have been deadly.

**Erica Mandy** (0:15)
Today is Saturday, August 1st. More and more people are turning to AI chat bots for health questions. And they're finding AI offers quick, confident answers. But how much should you trust what AI tells you? Today, we're hearing from two doctors with two important perspectives. Where AI can really help patients, and where the risks are still very real. We start with Dr. Matthew Callstrom from Mayo Clinic. There, he's the physician lead for the Artificial Intelligence Program, which is working with Microsoft to develop an AI model designed specifically for health care. So coming up, when Dr. Callstrom thinks AI can be phenomenal for patients, and when he says you definitely still need a human doctor.
Welcome to The Newsworthy Special Edition Saturday. When we sit down with a different expert or celebrity every Saturday to talk about something in the news. Don't forget to tune in every Monday through Friday for our regular episodes where we provide all the day's news in less than 15 minutes. I'm Erica Mandy, it's now time for today's special edition Saturday.
Dr. Matt Callstrom, thank you so much for joining us here on The Newsworthy.

**Matthew Callstrom** (1:18)
Great to be with you.

**Erica Mandy** (1:19)
So first, how are Americans using AI for health questions today? And what do you know about how widespread it is?

**Matthew Callstrom** (1:26)
One in seven queries of the models are around health care. So it's one of the most common uses.

**Erica Mandy** (1:33)
Where can AI be useful for someone trying to better understand their health or an acute situation?

**Matthew Callstrom** (1:40)
Yeah, it's a great question. I think for general questions, they're very good. I think if you have an acute question, do you have sinusitis or symptoms of sinusitis? And does it think that from the symptoms that people offer, that that might be the diagnosis?
I think it's reasonable. You can get pretty good advice that would otherwise be hard to get to through Google or just a regular search. So I think those next level questions on chronic conditions are really helpful.

**Erica Mandy** (2:07)
Okay, so that's interesting. So you and other doctors are on board with people using AI chat bots for at least some things, and you find that they can be pretty accurate for some things at least.

**Matthew Callstrom** (2:20)
Yeah, I think anything that informs patients about their own health, I think is really helpful.
And if the answers that are provided give them some insight, I think that's incredibly helpful. Getting to those more complex scenarios, that's where you want to certainly learn and manage your own health so that you come in prepared, and you can have a more informed discussion with that provider.

**Erica Mandy** (2:45)
So let's talk about where you don't think AI should be used ever, and also just other risks that people need to be aware of.

**Matthew Callstrom** (2:52)
What you hope for is that, if someone has a question about a serious health issue, that they may get to some response from a model, but I hope they don't stop there. I hope that they go on and interact with the health care provider that can help them deal with that issue, whatever it is that they're trying to understand, and then get to the right therapy, right intervention if that's needed.

**Erica Mandy** (3:15)
So is that because you don't want somebody to trust it to the point where they aren't getting the treatment they need? Is there any other reason? Is it a concern of accuracy or anything like that?

**Matthew Callstrom** (3:27)
There's a lot of variables in terms of getting to the right diagnosis.
One is symptoms, so I think people can have conversations around that.
Then you get into the next step questions. Should I have imaging done? Should I have another lab done? What should I do to try and get to a diagnosis? That's where you want to go to that next level, working with the health care providers so that you can actually get to the right answer.

**Erica Mandy** (3:52)
Do you think it's okay to upload certain medical personal information, perhaps test results to better understand them through AI, or even pictures of a certain rash you have, for example?

**Matthew Callstrom** (4:03)
Yeah, the last one, loading up a rash, I think that's a really great application for AI to say, gosh, was that a tick bite picture and I should put on medication? That's super helpful. Models have got pretty good from a dermatological standpoint. I think the loading your own health care information, I think that's a personal choice right now in terms of what these models do with the data right now.

14 more minutes of transcript below

Feed this to your agent

Try it now — copy, paste, done:

curl -H "x-api-key: pt_demo" \
  https://spoken.md/transcripts/1000651996090

Works with Claude, ChatGPT, Cursor, and any agent that makes HTTP calls.

From $0.10 per transcript. No subscription. Credits never expire.

Using your own key:

curl -H "x-api-key: YOUR_KEY" \
  https://spoken.md/transcripts/1000779388691