TLDR;

I asked ChatGPT + Grok 3 about what to do to deal with my chronic asthma. The advice they gave me, the data sources they surfaced and their ability to answer questions just outperformed 18 years of GP visits on this issue.

I link a copy of my ChatGPT transcript at the end.

Medical Context

I have chronic asthma. 

I have been going to various GPs in Australia for 18 years to get prescription medication for it.

My primary treatment is a prophylactic inhaler, that I’m supposed to take 1 inhalation twice a day. Initially this was Seretide (Fluticasone propionate + Salmeterol), but a few years ago, I switched to Symbicort (budesonide + formoterol)).

I also have a Ventolin inhaler (Salbutamol) that is meant to be used for short-term relief when I experience an asthma attack.

Exercise-induced asthma

Since Covid, I've also started doing cardio exercise. Doing so has resulted in me realising that I also have exercise-induced asthma, where I get symptoms as a direct result of the exertion on my lungs from doing heavy cardio.

I have been self-managing the exercised-induced asthma using an inconsistent combination added doses of the prophylactic inhaler before exercise and prophylactic inhaler or Ventolin during exercise itself when experience systems.

18 Years of existing GPs

I've been seeing a variety of GPs, in a variety of locations. My patient history in Australia spans 18 years, seeing GPs spread across Kingsford, Sydney, Surry Hills/Darlinghurst and Hurstville. Importantly, my prophylactic inhaler is prescription-only, which means I need to visit a GP semi-regularly to get a new script for it.

All GPs bar one have provided me with a new script to get more doses of my inhaler with only cursory questioning like "How is your asthma? Are you ok?".

One GP did mention that they were surprised I was using a prophylactic inhaler for exercise-induced asthma and suggested I use Ventolin instead, but otherwise never explained why or how to use Ventolin to control exercise-induced asthma.

Only one other GP actually gave me meaningful advice and told me to switch my inhaler from Seretide to Symbicort. (I later looked up his advice, and discovered he was correct to advise me to do so).

ChatGPT

ChatGPT - Initial Prompt

At one point, I realised that I could ask ChatGPT (4o + Deep Reasoning) about my asthma. (Note: I pay for the ChatGPT Plus plan costing 20USD/month) 

I told it about my medications, my symptoms and then asked if I had appropriate treatment.

ChatGPT asked me some clarifying questions and then went off to do its Deep Research thing.

Interestingly, (and this will be important later) I mis-specified my medication in my initial instruction, claiming that I took "seretide inhaler (Budesonide / Formoterol)", when in fact I should have said "symbicort inhaler (Budesonide / Formoterol)".

ChatGPT - First answer

The good stuff

ChatGPT ended up surfacing a variety of search results, the most important of which was the 2016 British Thoracic Society's Official Guideline on the Management of Asthma - technically a 29 page quick reference guide PDF. I read both ChatGPTs results and the guideline linked.

This contained the first bombshell of the chat, which is that all patients with asthma should have a "written personalised asthma action plan and be supported by regular professional review". No GP in Australia has ever talked to me about any kind of formal asthma action plan, let alone a written, personalised one.

Secondly, the guideline explains the hierarchy of interventions for asthma control - the idea that there is a defined hierarchy of interventions with escalating dosages/treatments. Individuals are supposed to follow their written plans, assess if their asthma is under control, if not go up one level on the hierarchy of interventions to a stronger dose of medication/treatment, otherwise maintain the current level of intervention. Eventually, if under control, experiment with going down a level of intervention to try and find the lowest level of intervention required for asthma control. It also explained how to measure asthma control, including through the use of a peak exhalation flow (PEF) meter.

No GP has ever talked to me about what "controlled" asthma looks like, how to monitor my symptoms to determine if I need to go up or down a hierarchy of control, nor have they explained to me what going up or down the hierarchy of control entailed.

Thirdly, it indicated that short-acting beta-agonists (SABAs) (Ventolin being one) on top of regular asthma control as a way to prevent breakout attacks during exercise are the preferred treatment for controlling exercised-induced asthma. Again, no GP has ever mentioned that I could use a SABA prophylactically, that is to say, to take it before an expected asthma attack, rather than after symptoms arose.

Fourthly, ChatGPT explained that my existing practice, of mostly not taking medication and treating asthma attacks when they arose using a combination of ICS + LABAs + SABAs was not actually good. This is the first time someone has explained to me that the objective of asthma medications should be to maintain asthma control so that no breakout attacks occur at all, and that the fact that I was having occasional ad-hoc attacks was actually a sign that I needed to increase my baseline control medications to eliminate these attacks.

ChatGPT also included lots of other information, alternative treatments options including some that I recognise that I was on in childhood. All in all, it provided an extremely good baseline of what I should do for sure, and what else I might want to do.

One Drawback to highlight

One interesting drawback - ChatGPT got confused by my mistake - mixing up Seretide and Symbicort. Although this did not have any material consequence to it's recommendations (since Seretide and Symbicort are both ICSes with LABAs), it reproduced my mistake, by referring to Seretide as "Budesonide / Formoterol", instead of picking up that Seretide is in fact "Fluticasone propionate + Salmeterol", or noting that "Budesonide / Formoterol" is Symbicort. So ChatGPT wasn't capable of picking up my error, admittedly a relatively inconsequential one in this case.

ChatGPT - Follow ups

I then asked ChatGPT to write me a personalised asthma action plan, as recommended by the earlier sources. It did so without issue. 

I also got the chance to ask follow up questions relating to things in the asthma action plan.

For example, I had to ask ChatGPT to explain to me how to use a PEF meter to calibrate my exhalation performance therefore how to assess if my asthma was under control. This is the first time anyone/anything has talked to me about how to assess my own medical condition and how to determine if/when I need to seek help from a doctor to adjust my medication upwards.

The instructions were unclear, so I asked ChatGPT to explain to me how often I should be using my PEF meter (Answer: daily). Again, no one had ever talked to me about frequency of measurement for my personal monitoring or how to track this over time. More than that though, the ability to ask clarifying questions meant that aspects of the guideline that would normally have been implied through context known to a doctor but not a patient, could now be supplied by an LLM.

Grok 3 - Alternative LLM

After the very positive experience with ChatGPT, I decided to feed a similar prompt into Grok 3. I use the free version of Grok with its reasoning equivalent. 

Grok - Initial Result

Grok had response that was comparable to ChatGPT. This is of itself very impressive. 

Grok also picked up my mistake in describing my medication, and despite otherwise using my nomenclature, gently mentioned that I might be wrong. Specifically, it added a line, "However, your use of Seretide (assumed to be Budesonide/Formoterol based on your description, though typically Seretide is Fluticasone/Salmeterol) prior to exercise, rather than daily, raises questions." Again, very impressive.

Grok based its results on the Australian National Asthma Council's Australian Asthma Handbook (the Australian equivalent of the British Thoracic Society Guidelines). I suspect it used my IP address/other metadata and knew to personalise my results to Australia - which is also very impressive.

The guidelines are broadly similar, with the Australian guidelines advising that all patients should have a "written asthma action plan (WAAP)".

Synthesis

I then got both ChatGPT and Grok 3 a link to the Australian guidelines, and ask them to produce a WAAP. In the end, I found the ChatGPT one to be a little nicer to read and deal with and so I've  been using that as my WAAP. 

Impact on my actual health

Importantly, that's just words on a screen. Let's talk a bit about my actual health.

I immediately started following this new written plan. I bought a PEF meter on Amazon (free next day shipping with Prime!), I start tracking my PEF meter readings and taking my prophylactic inhalations regularly - logging everything in my phone.

Things work great! My lung capacity as measured by my PEF is now notably higher than when I started. I no longer have uncontrolled asthma attacks. When I maintain my exercise regimen, my exercise induced asthma is actually controllable with fewer Ventolin inhalations.

Extension

Follow up questions as a result of following the plan

Another fun thing is the ability to ask tiny follow questions that you'd never be able to with a doctor. After using the WAAP for a few weeks, my PEF actually increased. I then asked ChatGPT follow up questions, on when and how I should re-baseline my PEF and whether or not the new baseline readings would make sense given that they're in part influenced by my use of medicines.

Final Notes

So there you have it. ChatGPT + Grok has outperformed the collection of GPs I've seen over the last 18 years. My asthma symptoms have literally never been better.

A link to my ChatGPT transcript is available here.