Targeting Immune Aging: New Findings from the MitoImmune Study on Urolithin A (Mitopure®)
An excellent colleague of mine recently had an unpleasant experience when she discovered that every single citation in an article she’d generated using AI and sent over as a potential guest post appeared fine on first glance, but was completely fabricated. Not one existed. Thankfully, our team caught it very quickly and never would have approved it for publication without checking. But I’ll admit: I understand the temptation. Most of us are balancing a dozen spinning plates. If AI can lighten the load, who wouldn’t be leaning in?
“Intelligence” everywhere…
2025 feels like the year that AI seeped into every nook and cranny. None of us can do a Google search anymore without an AI summary perched at the top. According to the University of Virginia Darden School of Business, nearly 60 percent of consumers now use AI to help them shop, and more than half of us would rather ask AI what to wear than ask a friend.
Small businesses – many of us functional medicine clinicians among them – are on the adoption trend, too. A recent Business Wire’s report found 92 percent of small businesses have integrated AI into their operations in some form, up from just 20 percent two years ago. It’s described as a “productivity multiplier” both in process automation and in research/analysis tasks and even though there’s likely some overhype of its current potential, AI is here to stay. In fact, there’s really only one viable trajectory: AI will never be as bad as it is today.
But what about functional and longevity medicine? I’ll admit that, even as AI is creeping judiciously into our own operational model (though never left to its own devices!), there are ongoing questions: What’s the best way to use AI in functional and longevity medicine specifically? What new capabilities are emerging? What tools should we choose? What is hype vs reality?
That’s exactly what we set out to explore in our 2025 Practitioner Survey on The Use of AI and Wearables in Functional and Longevity Medicine. We asked our community how they’re using these tools, what they’re learning, and where they’re hitting roadblocks. The resulting report is grounded in real clinician experience – your experience. By sharing this kind of knowledge across our community, I believe we will also benefit and help each other.
I’m bullish on AI, but with caveats
I think there is value in experimenting with AI tools to see where we can improve both effectiveness and efficiency. Honestly, some of the output is just mindblowing! You can put me in the “pro-AI” camp. BUT, with a hefty dose of caution. So far, using AI hasn’t exactly been seamless.
In a recent case of pediatric systemic lupus erythematosus (SLE), I had a positive experience integrating AI into my work. I suspected that gluten sensitivity could be contributing to immune dysregulation for my patient. The patient’s father, an ophthalmologist, needed to see the scientific connections before proceeding, so I used AI to surface relevant data on the connections between HLA-DQ2/DQ8 genotypes, antigen presentation, and intestinal permeability. These alleles, while classic in celiac disease, also appear more broadly across autoimmune conditions, linking aberrant MHC II binding with mucosal immune activation.
Based on this, the father could understand why I wanted to do haplotype testing. In less than the blink of an eye, AI added drug-nutrient depletions that were relevant to her case. It also pulled references for n-acetylcysteine (NAC) pediatric dosing; there were many in the literature, which provided a more nuanced input than just using Clark’s Rule alone, since we can see how aggressively/conservatively it’s been used. I had this patient on NAC because her liver function tests had been high (possibly exacerbated by medications). However, there’s also early clinical and preclinical data for its use in offsetting the increased mitochondrial oxidative stress seen in SLE.
While the AI handled literature retrieval and synthesis, as well as some draft recommendations, in seconds, work that previously required my own manual searching online and sometimes in my wall of textbooks, the clinical reasoning, along with ownership of the accuracy of the output, remained mine. The tool accelerated the process and surfaced some information that I might not have connected, but everything still required verification.
Still, the time savings aren’t always apparent. There are other cases (let’s say the even more complex/unique, which are not uncommon in my practice) in which AI has definitely bumped up my ability to practice deep functional medicine. However, this seems to come at the cost of added complexity to my workload. AI lab integrations remain frustratingly absent as well, at least in my experience. And we clinicians need help with the volume of lab and other data, such as from wearables.
We asked about wearables, too!
I admit, I love my Oura data (and I have no commercial affiliation). I also want to look at Whoop, since I noticed that several of you report using it. But, just to underline the challenge: patients frequently send me screenshots of their Oura data, which requires extra manual steps to integrate into an EHR.
Of course, we need to be able to receive and integrate that information, but there’s a bit of a logistical challenge getting that into a chart-ready format, let alone something that we can put into a long-term observational system. Based on a useful tip in the Survey Report, I’m now uploading them to AI to extract the data from those images, which is an improvement. Here’s a recent example of what AI came up with from a couple of Oura screenshots:
I’ve also been pretty happy with the Freestyle Libre continuous glucose monitor (CGM) but recently wanted to test out the Stelo based on Dr. Rountree’s mention of its Oura integration in my recent Masterclass (replays available within the Younger You practitioner training). I popped one on the other day and set off on my routine bike ride, but was in for a surprise: the CGM registered my glucose as off the charts high!
I was taken aback since I’ve never had that high a reading in many CGM uses, but decided to persist. Over the next few days, it seemed to track at least 10 points above where I’d expect, and then, another surprise, it started to normalize down to ranges that I’m more familiar with seeing.
Asking around, I found that others had experienced this same pattern. One bit of advice that came my way was to put the CGM on two or more hours after your last meal of the day so that its first readings are overnight. Perhaps some of you have also had experience with Stelo and would like to share your comments and tips below.
My takeaways, though, are that we can’t expect supreme accuracy with some tools, and especially with any one measure. As Dr. Rountree reminded us, it’s essential to look at readings over time and in context, and not place excessive emphasis on any one value. Incidentally, and highly reflective of the journey we’re on right now, the Stelo and my Oura ring were never able to integrate. Unless this bug can be fixed, there is no reason to deviate from Freestyle.
Download your copy of the report and let us know what you think
Everyone who kindly completed our survey will receive a complimentary copy of the report, as our thanks. For anyone else who would like to find out how your peers in functional and longevity medicine are using AI and wearables, you can get your copy here for a nominal fee (intended to cover the cost of production).
If you’d like to share your thoughts on this topic, including your own experience with AI, please add a comment below!
Related posts
