Can AI take over the bulk of what General Practitioners do?

mraajr

mraajr

Explorer
Joined
Oct 14, 2025
Messages
10
Reaction score
0
Location
Missouri
I never imagined I would be posing this, but over the past few days, after discussing my latest bloodwork with Claude.AI, it seems obvious to me that AI handling a large share of GP duties is bound to happen. The fast replies, precision, care plans, and so on have all proven incredibly helpful. I have been far more impressed than expected and will definitely keep turning to it for quick guidance.

Naturally, I believe something like medicine will forever (or at least for the foreseeable future) require a human component. People from older generations, raised long before the internet era, will find it tougher to adapt. That is not a criticism, just a largely instinctive response. Yet younger people who grow up surrounded by this technology will almost certainly have far less difficulty accepting medical guidance from a computer.

Might AI eventually take over most GP-style roles? Might insurers start using AI to remove them from the process entirely and issue direct specialist referrals when necessary? Might it even reach a stage where it can prescribe certain medications? Or are there other applications in this setting that I have not considered?
 
My answer would be yes. When you go to most family practice doctors, what happens is that they take the symptom list you bring in and match it, without much critical thought, against the standard of care treatment, while also ordering the labwork and vaccine schedule that is recommended for your age range.

That said, some family practice doctors really are skilled at using judgement and discernment to tailor these things to the individual. Running into one of those, however, is a welcome surprise you weren't counting on.
 
The other day I asked ChatGPT which wood to use for a brisket — post oak or hickory. My assumption was that it would look through smoking forums and similar online sources, then report back on which option had stronger backing.

It ended up giving me a "Which response do you prefer" prompt, showing the same content in two different styles — something it occasionally does. The information WAS laid out in two distinct ways, and beyond that, one version came down firmly on hickory while the other SWORE by post oak.
 
randompersonrandom said:

The other day I asked ChatGPT which wood to use for a brisket — post oak or hickory. My assumption was that it would look through smoking forums and similar online sources, then report back on which option had stronger backing.

It ended up giving me a "Which response do you prefer" prompt, showing the same content in two different styles — something it occasionally does. The information WAS laid out in two distinct ways, and beyond that, one version came down firmly on hickory while the other SWORE by post oak.
Click to expand...
What you're posing to it is a matter of opinion, not a matter of fact. Admittedly, in the end medicine does come down to opinion, yet the majority of physicians approach it as though it were factual and take the guidelines handed to them as the final authority, which cuts their value down substantially.
 
mraajr said:

I never imagined I would be posing this, but over the past few days, after discussing my latest bloodwork with Claude.AI, it seems obvious to me that AI handling a large share of GP duties is bound to happen. The fast replies, precision, care plans, and so on have all proven incredibly helpful. I have been far more impressed than expected and will definitely keep turning to it for quick guidance.

Naturally, I believe something like medicine will forever (or at least for the foreseeable future) require a human component. People from older generations, raised long before the internet era, will find it tougher to adapt. That is not a criticism, just a largely instinctive response. Yet younger people who grow up surrounded by this technology will almost certainly have far less difficulty accepting medical guidance from a computer.

Might AI eventually take over most GP-style roles? Might insurers start using AI to remove them from the process entirely and issue direct specialist referrals when necessary? Might it even reach a stage where it can prescribe certain medications? Or are there other applications in this setting that I have not considered?
Click to expand...

Why stop at primary care? Consider the bigger picture. Imaging scans are already being read by AI, illnesses diagnosed, treatment plans put together, and so on.

When it comes to handling my sleep apnea treatment, AI outperforms the specialist office I go to — and that office even has a whole team of respiratory therapists. I take the reports my cpap machine generates, send them to AI, and it reviews them, gives feedback, and suggests changes to my cpap settings. Meanwhile, all I hear from my doctor's office is how well my treatment is working 🙄.

On the referral point you raised — that shift is already underway. I came across an article saying that classic medicare beneficiaries will now need AI approval for prior authorization on 17 procedures. No industry is safe from AI.
 
Turning to chatgpt (or comparable AI tools) for medical guidance, unless it is checked against something outside the model — ideally a physician — carries real risk.

Right now the main issue with chatgpt in a medical context is hallucination. Physicians can likewise state things with confidence that turn out to be wrong, yet chatgpt is more prone to that, while doctors are far more inclined to request additional testing, send you to a specialist, or occasionally simply hold off and check again later to observe how a condition evolves when they are unsure what direction to take. This happens often; plenty of medical problems never yield a clear diagnosis.

I have leaned on it heavily for questions about medications and as an aid while trying to figure out a rash that was hard to diagnose. Months went by, along with blood tests, 3 biopsies, my gp plus 2 dermatologists and a few dermatology residents, and an enormous amount of reading and reviewing dermatology images — genuinely hundreds of hours of research. Eventually the specialists gave me a vague answer that was nearly the same as what I and chatgpt had arrived at. That actually felt reassuring to me, even with no specific diagnosis and no treatment options.

One thing I noticed: the wording mattered a great deal. Using proper medical terminology affects how good the answers are, and the way the problem is framed is hugely important — posing the same question differently produces different replies. In my profile I told it to flag when my statements were incorrect, which is likely essential to keep it from endorsing false assumptions; absent that, its inclination to agree with you could become a serious problem.

On knowledge, it is impressive, and in my view — supported by some research papers — it comes fairly close to specialist level, in certain areas with the right prompts even surpassing that, and clearly ahead of a gp by a wide margin.

On logic, chatgpt is honestly still not very strong, and weaker than most people, so wherever logical reasoning is required it may commit significant errors. That is not always necessary for medical advice, but at times it is essential.

On clinical perspective and judging what matters and how to approach a problem, it is unexpectedly quite capable, though still usually below medical advice.

Still, being more or less free and reachable at any hour with no waiting is a big plus, and it widens access to expert-level advice, and it is particularly strong at identifying rarer conditions.
 
I rely on chatgpt as a tool for understanding and managing my peptide protocol and my weight loss. Everything gets fed into it: my prescriptions, all my labs and bloodwork, plus the details of my day-to-day life, from my job to whatever my weight happens to be.

It has a REALLY thorough picture of who I am. On top of that, it doubles as medical advice, since I have no intention of telling my dr. that I split-dose reta and tirz, that I take 2 daily shots of a wolverine stack, or that once my ipamorelin and HGH arrive, I might try stacking those as well. Chat GPT walks me through all of it. Sure, it gets things wrong here and there, and that does not bother me.

There is no judgment from it, and it keeps helping me with whatever I am dealing with. At the moment, that means shooting chems from unmarked vials sold by strangers on the internet into my body several times each day.

So yes. My guess is that chatgpt will take over the role of most general practitioners, and in time the specialists too.
 
Since medical errors rank as the 3rd top reason people die in the US, my trust leans toward AI. During my time employed at a healthcare facility, I saw Nurses griping instead of doing their duties, Doctors prioritizing their automobiles over those they treated, administrators fixated solely on their paychecks, and everyone was hooking up with each other.
 
AI is already making inroads into healthcare, including general practice.

Last year I had a follow-up appointment for an illness. Since my usual physician was unavailable, I ended up seeing a different doctor at the same clinic.

Out of nowhere, he asked if he could record our conversation on his phone to transcribe it.

I wasn't comfortable with that, but I didn't want things to get awkward with a doctor I'd just met, so I agreed.

From what I could tell, his phone—no idea which app—picked up everything we said and transcribed it. Then it produced a summary, and somehow it was linked to the computer in each exam room that's hooked up to their network.

It didn't look like he was doing much thinking himself; the app summarized it all, and then he began consulting the room's computer terminal to check treatments and medications.

Whenever I had a specific question, he'd type it into the computer and wait for it to tell him how to respond and what to suggest.

How is that different from an ordinary person turning to chat-gpt or grok, etc., for advice?

Some might argue that at an actual doctor's office, you at least have a real doctor reviewing what the AI says.. but from what I've seen, some doctors don't appear to do that. They seem to depend almost completely on AI to listen, take notes, evaluate, and make recommendations.

And I'm expected to pay exorbitant premiums and visit fees so someone can use AI to 'treat' me? 🤔😔
 
Sure, a few physicians are terrible, and Claude would outperform them. That said, genuinely skilled doctors still exist, and they'll notice details through direct observation and their own reasoning that never made it into a Claude conversation—that kind of expertise still matters. On top of that, LLMs continue to make things up, and even as that improves, the fabrications arrive with complete certainty and are extremely convincing. Whenever Claude gives me something, I verify it against conventional sources before trusting the analysis.
 
Grogu said:

mraajr said:

I never imagined I would be posing this, but over the past few days, after discussing my latest bloodwork with Claude.AI, it seems obvious to me that AI handling a large share of GP duties is bound to happen. The fast replies, precision, care plans, and so on have all proven incredibly helpful. I have been far more impressed than expected and will definitely keep turning to it for quick guidance.

Naturally, I believe something like medicine will forever (or at least for the foreseeable future) require a human component. People from older generations, raised long before the internet era, will find it tougher to adapt. That is not a criticism, just a largely instinctive response. Yet younger people who grow up surrounded by this technology will almost certainly have far less difficulty accepting medical guidance from a computer.

Might AI eventually take over most GP-style roles? Might insurers start using AI to remove them from the process entirely and issue direct specialist referrals when necessary? Might it even reach a stage where it can prescribe certain medications? Or are there other applications in this setting that I have not considered?
Click to expand...

Why stop at primary care? Consider the bigger picture. Imaging scans are already being read by AI, illnesses diagnosed, treatment plans put together, and so on.

When it comes to handling my sleep apnea treatment, AI outperforms the specialist office I go to — and that office even has a whole team of respiratory therapists. I take the reports my cpap machine generates, send them to AI, and it reviews them, gives feedback, and suggests changes to my cpap settings. Meanwhile, all I hear from my doctor's office is how well my treatment is working 🙄.

On the referral point you raised — that shift is already underway. I came across an article saying that classic medicare beneficiaries will now need AI approval for prior authorization on 17 procedures. No industry is safe from AI.
Click to expand...
Yeah, AI is already handling that. The way I first got put on Mounjaro was because my blood glucose reading was the highest you can have without being diabetic. I use glasses — nearsighted, with bifocals for reading — but mine are progressive lenses, so they cover the full range of focal lengths from distance to close up. About 2 months after the diagnosis that said I was roughly 15 minutes away from becoming diabetic (and while I was still waiting on a list for the starting dose of Mounjaro), I managed to get in for an eye exam (with the time restriction on insurance). I figured I just needed a stronger prescription since reading had gotten harder. Then the eye doctor tells me my vision had actually IMPROVED by 2 whole diopters over the past year. That's not typical, and it usually points to elevated blood sugar, because that makes the lenses in your eyes swell and change shape, so farsighted people become less farsighted. I said oh my god, I literally just got that diagnosis! So my eye doctor said the standard of care is to photograph your retina and check for diabetic retinopathy. We did that, and now AI software analyzes the image to spot it. My doctor told me she'd look too, since she doesn't fully trust it, but it hasn't been wrong once since she started using it. Good news: there was no sign of it. And I got new glasses. Also, as my blood sugar returned to normal over the next year, your lenses reshape back to how they were, and my vision got worse again.
 
ocuzuc said:

Grogu said:

mraajr said:

I never imagined I would be posing this, but over the past few days, after discussing my latest bloodwork with Claude.AI, it seems obvious to me that AI handling a large share of GP duties is bound to happen. The fast replies, precision, care plans, and so on have all proven incredibly helpful. I have been far more impressed than expected and will definitely keep turning to it for quick guidance.

Naturally, I believe something like medicine will forever (or at least for the foreseeable future) require a human component. People from older generations, raised long before the internet era, will find it tougher to adapt. That is not a criticism, just a largely instinctive response. Yet younger people who grow up surrounded by this technology will almost certainly have far less difficulty accepting medical guidance from a computer.

Might AI eventually take over most GP-style roles? Might insurers start using AI to remove them from the process entirely and issue direct specialist referrals when necessary? Might it even reach a stage where it can prescribe certain medications? Or are there other applications in this setting that I have not considered?
Click to expand...

Why stop at primary care? Consider the bigger picture. Imaging scans are already being read by AI, illnesses diagnosed, treatment plans put together, and so on.

When it comes to handling my sleep apnea treatment, AI outperforms the specialist office I go to — and that office even has a whole team of respiratory therapists. I take the reports my cpap machine generates, send them to AI, and it reviews them, gives feedback, and suggests changes to my cpap settings. Meanwhile, all I hear from my doctor's office is how well my treatment is working 🙄.

On the referral point you raised — that shift is already underway. I came across an article saying that classic medicare beneficiaries will now need AI approval for prior authorization on 17 procedures. No industry is safe from AI.
Click to expand...
Yeah, AI is already handling that. The way I first got put on Mounjaro was because my blood glucose reading was the highest you can have without being diabetic. I use glasses — nearsighted, with bifocals for reading — but mine are progressive lenses, so they cover the full range of focal lengths from distance to close up. About 2 months after the diagnosis that said I was roughly 15 minutes away from becoming diabetic (and while I was still waiting on a list for the starting dose of Mounjaro), I managed to get in for an eye exam (with the time restriction on insurance). I figured I just needed a stronger prescription since reading had gotten harder. Then the eye doctor tells me my vision had actually IMPROVED by 2 whole diopters over the past year. That's not typical, and it usually points to elevated blood sugar, because that makes the lenses in your eyes swell and change shape, so farsighted people become less farsighted. I said oh my god, I literally just got that diagnosis! So my eye doctor said the standard of care is to photograph your retina and check for diabetic retinopathy. We did that, and now AI software analyzes the image to spot it. My doctor told me she'd look too, since she doesn't fully trust it, but it hasn't been wrong once since she started using it. Good news: there was no sign of it. And I got new glasses. Also, as my blood sugar returned to normal over the next year, your lenses reshape back to how they were, and my vision got worse again.
Click to expand...
Were you certain your levels weren't already in the diabetic range prior to starting Mounjaro? My own diabetes discovery followed a very similar path. I was on a trip and realized I could no longer make out the menu boards at the airport when trying to order, so I started snapping photos of them with my phone and reading the food options off the screen. Just like you, my initial assumption was that vision naturally shifts once you hit your 40s, so I got glasses, which fixed the problem, and then booked a checkup because the eye doctor mentioned diabetes as a possible underlying cause.

Ahead of that visit, I bought a blood sugar meter, and it turned out I was sitting around 600 mg/dL. The very first result the device gave me was simply "HI," which made me think the thing was malfunctioning. The next test came back at just under 600. My girlfriend measured 80, so the machine clearly wasn't the problem. I suppose "HI" was the meter's way of saying my reading was beyond what it could register. LOL
 
There's a Jama study worth looking at here. On its own, AI reached a 92% accurate diagnosis rate. Physicians working alone hit 72%. When doctors had AI assistance, accuracy came in at just 74%. "An unexpected secondary result was that the LLM alone performed significantly better than both groups of humans, similar to a recent study with different LLM technology." https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2825395

This is something I've been talking about for ages, though I won't dump a giant wall of text on everyone. Broadly speaking, the U.S. medical system is broken. From testing and approving new devices and medicines, through diagnosis and treatment plans, to cutting out unnecessary tests, surgeries, follow up care, and insurance processing, AI could reduce costs and improve every stage. I had Gemini AI put together a detailed plan with savings estimates. Its conclusion: over 10 years the savings would reach $1.2 trillion dollars, and healthcare outcomes would improve dramatically.

Pouring ever more money into a broken medical system won't repair it. Technology adoption will -- that is, if the entrenched money interests don't stand in the way of progress.
 
That arrangement persists for a purpose, though. Instead of anything close to a competitive free market system, we get government gatekeeping, regulations, and controls because people irrationally pursue what they perceive as the "best" care and quality control. What we're left with is an inefficient, self-defeating, corrupt, and overpriced monster, mostly made possible by emotional thinking.

Fixing this never required advanced machine learning or AI. Decades ago, simple if/then algorithms were capable of replacing most doctor functions. The obstacle wasn't a lack of sophistication, but entrenched interests and inertia. So your worry about that acting as a barrier seems right to me.
 
tubby said:

That arrangement persists for a purpose, though. Instead of anything close to a competitive free market system, we get government gatekeeping, regulations, and controls because people irrationally pursue what they perceive as the "best" care and quality control. What we're left with is an inefficient, self-defeating, corrupt, and overpriced monster, mostly made possible by emotional thinking.

Fixing this never required advanced machine learning or AI. Decades ago, simple if/then algorithms were capable of replacing most doctor functions. The obstacle wasn't a lack of sophistication, but entrenched interests and inertia. So your worry about that acting as a barrier seems right to me.
Click to expand...
On a separate note, this is impossible regardless. With the exception of genuinely optional treatments, health care stands as the textbook example of a market with inelastic demand.
 
Here's my 2 cents: AI is overblown. It doesn't make good sense, and its information sources are often way outdated. It doesn't bring human thought into the equation. If a drug had a 5% death rate after first dose, would you take it? AI says 95% chance of success, that's good. For a runny nose, hell no. For a 100% fatal cancer, absolutely yes.
 
tubby said:

ocuzuc said:

Grogu said:

mraajr said:

I never imagined I would be posing this, but over the past few days, after discussing my latest bloodwork with Claude.AI, it seems obvious to me that AI handling a large share of GP duties is bound to happen. The fast replies, precision, care plans, and so on have all proven incredibly helpful. I have been far more impressed than expected and will definitely keep turning to it for quick guidance.

Naturally, I believe something like medicine will forever (or at least for the foreseeable future) require a human component. People from older generations, raised long before the internet era, will find it tougher to adapt. That is not a criticism, just a largely instinctive response. Yet younger people who grow up surrounded by this technology will almost certainly have far less difficulty accepting medical guidance from a computer.

Might AI eventually take over most GP-style roles? Might insurers start using AI to remove them from the process entirely and issue direct specialist referrals when necessary? Might it even reach a stage where it can prescribe certain medications? Or are there other applications in this setting that I have not considered?
Click to expand...

Why stop at primary care? Consider the bigger picture. Imaging scans are already being read by AI, illnesses diagnosed, treatment plans put together, and so on.

When it comes to handling my sleep apnea treatment, AI outperforms the specialist office I go to — and that office even has a whole team of respiratory therapists. I take the reports my cpap machine generates, send them to AI, and it reviews them, gives feedback, and suggests changes to my cpap settings. Meanwhile, all I hear from my doctor's office is how well my treatment is working 🙄.

On the referral point you raised — that shift is already underway. I came across an article saying that classic medicare beneficiaries will now need AI approval for prior authorization on 17 procedures. No industry is safe from AI.
Click to expand...
Yeah, AI is already handling that. The way I first got put on Mounjaro was because my blood glucose reading was the highest you can have without being diabetic. I use glasses — nearsighted, with bifocals for reading — but mine are progressive lenses, so they cover the full range of focal lengths from distance to close up. About 2 months after the diagnosis that said I was roughly 15 minutes away from becoming diabetic (and while I was still waiting on a list for the starting dose of Mounjaro), I managed to get in for an eye exam (with the time restriction on insurance). I figured I just needed a stronger prescription since reading had gotten harder. Then the eye doctor tells me my vision had actually IMPROVED by 2 whole diopters over the past year. That's not typical, and it usually points to elevated blood sugar, because that makes the lenses in your eyes swell and change shape, so farsighted people become less farsighted. I said oh my god, I literally just got that diagnosis! So my eye doctor said the standard of care is to photograph your retina and check for diabetic retinopathy. We did that, and now AI software analyzes the image to spot it. My doctor told me she'd look too, since she doesn't fully trust it, but it hasn't been wrong once since she started using it. Good news: there was no sign of it. And I got new glasses. Also, as my blood sugar returned to normal over the next year, your lenses reshape back to how they were, and my vision got worse again.
Click to expand...
Were you certain your levels weren't already in the diabetic range prior to starting Mounjaro? My own diabetes discovery followed a very similar path. I was on a trip and realized I could no longer make out the menu boards at the airport when trying to order, so I started snapping photos of them with my phone and reading the food options off the screen. Just like you, my initial assumption was that vision naturally shifts once you hit your 40s, so I got glasses, which fixed the problem, and then booked a checkup because the eye doctor mentioned diabetes as a possible underlying cause.

Ahead of that visit, I bought a blood sugar meter, and it turned out I was sitting around 600 mg/dL. The very first result the device gave me was simply "HI," which made me think the thing was malfunctioning. The next test came back at just under 600. My girlfriend measured 80, so the machine clearly wasn't the problem. I suppose "HI" was the meter's way of saying my reading was beyond what it could register. LOL
Click to expand...

Yes. I originally sought medical attention due to frequent urination and a constant urge to go. After a fasting blood draw, my glucose level was 137 and my A1c was 6.4. The doctor told me: “good news, you’re not diabetic, but you will be in about 15 minutes.” He then showed me a Lilly report indicating that tirzepatide had a 94% cure rate for pre-diabetes, and remarked, “this is unprecedented in medicine.” I had never heard of it before.

However, this occurred during the shortage, so I was placed on a waiting list.
 
tubby said:

randompersonrandom said:

The other day I asked ChatGPT which wood to use for a brisket — post oak or hickory. My assumption was that it would look through smoking forums and similar online sources, then report back on which option had stronger backing.

It ended up giving me a "Which response do you prefer" prompt, showing the same content in two different styles — something it occasionally does. The information WAS laid out in two distinct ways, and beyond that, one version came down firmly on hickory while the other SWORE by post oak.
Click to expand...
What you're posing to it is a matter of opinion, not a matter of fact. Admittedly, in the end medicine does come down to opinion, yet the majority of physicians approach it as though it were factual and take the guidelines handed to them as the final authority, which cuts their value down substantially.
Click to expand...
That's exactly why people say you should "get a second opinion" when your current doctor's advice doesn't sit right with you.

IshimaruKenta said:

No. AI can eat a dick.
Click to expand...
😂😂😂

Just wait — it'll swallow them all up soon enough!

m100568 said:

There's a Jama study worth looking at here. On its own, AI reached a 92% accurate diagnosis rate. Physicians working alone hit 72%. When doctors had AI assistance, accuracy came in at just 74%. "An unexpected secondary result was that the LLM alone performed significantly better than both groups of humans, similar to a recent study with different LLM technology." https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2825395

This is something I've been talking about for ages, though I won't dump a giant wall of text on everyone. Broadly speaking, the U.S. medical system is broken. From testing and approving new devices and medicines, through diagnosis and treatment plans, to cutting out unnecessary tests, surgeries, follow up care, and insurance processing, AI could reduce costs and improve every stage. I had Gemini AI put together a detailed plan with savings estimates. Its conclusion: over 10 years the savings would reach $1.2 trillion dollars, and healthcare outcomes would improve dramatically.

Pouring ever more money into a broken medical system won't repair it. Technology adoption will -- that is, if the entrenched money interests don't stand in the way of progress.
Click to expand...
The whole thing is broken by design. AI is built to seek answers and gets rewarded for being right. It has no awareness of whether it's actually correct. All it does is follow an extremely sensitive algorithm. There's no other incentive driving it — unlike humans, sadly.
 
Gr33dyOctopus said:

I rely on chatgpt as a tool for understanding and managing my peptide protocol and my weight loss. Everything gets fed into it: my prescriptions, all my labs and bloodwork, plus the details of my day-to-day life, from my job to whatever my weight happens to be.

It has a REALLY thorough picture of who I am. On top of that, it doubles as medical advice, since I have no intention of telling my dr. that I split-dose reta and tirz, that I take 2 daily shots of a wolverine stack, or that once my ipamorelin and HGH arrive, I might try stacking those as well. Chat GPT walks me through all of it. Sure, it gets things wrong here and there, and that does not bother me.

There is no judgment from it, and it keeps helping me with whatever I am dealing with. At the moment, that means shooting chems from unmarked vials sold by strangers on the internet into my body several times each day.

So yes. My guess is that chatgpt will take over the role of most general practitioners, and in time the specialists too.
Click to expand...
Yeah, that whole being-judged thing. Mine won't touch compounding pharmacies at all — he refused even while I spent months on a waitlist during the shortage. Grey market? Absolutely not. And the last thing I want is a line in my chart reading “patient is using internet purchased non pharmaceutical tirzeptide of unknown provence against recommendation.” Sure, in theory it's confidential, but to get insurance to cover the doctor (aside from the office visit copay) you generally waive your HIPAA rights anyway. If he found out what I've actually been doing, I doubt he'd lift a finger to help. As far as he knows, I'm still taking the Mounjaro he prescribes. I'm not, though, since my insurance stopped covering it.
 
Back
Top