Welcome back to Healthy Innovations! 👋
The AI health assistant space moves fast, and it can be hard to keep up with weekly announcements from both big tech and smaller startups. Today, I want to make sense of the latest moves from the former: four companies – two AI-native platforms and two long-standing tech giants – each with a distinct strategy for capturing this growing market.
Let’s dive in!
What is an AI health assistant?
You have probably already used one – typing symptoms into ChatGPT at midnight, uploading a blood test, or asking what a diagnosis actually means. An AI health assistant turns that ad‑hoc behaviour into a product. It can pull in your records, lab results, wearable signals, and health history, then use AI to help you understand what it all means and what to do next. The key difference from a Google search is context: it is designed to reason over your information, not just the open internet.
In the first three months of 2026, the space shifted fast.
OpenAI, Microsoft, Amazon, and Perplexity each launched a dedicated consumer health AI product in rapid succession – and each is making a very different bet on how a health assistant should work.
OpenAI: The front-door play
OpenAI moved first and most visibly. In January, it launched ChatGPT Health for consumers and OpenAI for Healthcare for enterprise – in the same week.
ChatGPT Health gives users a dedicated space within ChatGPT to connect medical records, lab results, and wellness apps, with health conversations kept separate from other chats and excluded from model training. Over 230 million people globally were already asking ChatGPT health questions every week. OpenAI is now building a proper home for that behavior.
The enterprise product is already live at major US health systems including Cedars-Sinai, Memorial Sloan Kettering, and UCSF, drawing on peer-reviewed research with direct citations.
What OpenAI is betting on: Owning the first interaction – for both consumers and health systems. The patient who uploads a lab result and the clinician who needs a summary at the point of care go to the same platform.
The gaps to watch: ChatGPT Health is not available in the European Economic Area, Switzerland, or the UK, showing how privacy rules can still block global rollout. A deeper question is whether these tools work as advertised. A randomised controlled study, undertaken at Oxford and published in Nature Medicine, tested 1,298 UK participants using GPT-4o and two other models across ten medical scenarios. While models alone identified relevant conditions in up to 94.9% of cases, participants using them scored under 34.5% – no better than standard internet search. Researchers argued the issue was human-AI interaction, and that benchmark scores correlated poorly with real-world outcomes. The study tested general-purpose models, so whether specialised products improve results at scale remains unclear.
Microsoft: The trust play
Microsoft launched Copilot Health on March 12. Where OpenAI leads with scale, Microsoft leads with institutional credibility.
Copilot Health connects to health records from more than 50,000 healthcare organizations or provider networks, data from over 50 wearable devices, and real-time provider directories, with expert-written content verified by a clinical team drawing on sources across 50 countries. Microsoft's Copilot was already fielding more than 50 million health questions a day before the dedicated product launched.
What Microsoft is betting on: Trust built over decades in enterprise healthcare. As Microsoft AI head Mustafa Suleiman put it: "We are a trusted brand, because Microsoft is old and wise, stable and committed for the very long term."
The gap to watch: Copilot Health launched in the US only. UK expansion is contingent on the MHRA's new regulatory framework for AI in healthcare, expected later in 2026.
Amazon: The vertical play
Amazon built something structurally different. Health AI – expanded from One Medical to the broader Amazon website and app this March – doesn't just answer health questions. It can explain lab results, renew prescriptions, and book clinical appointments, handing off directly to a One Medical clinician when the situation calls for it. Everything stays in the same app.
What Amazon is betting on: A closed loop connecting AI, pharmacy, and clinical care in a single experience – with a real provider network behind it. As Amazon's chief medical officer put it: "Ours is the only solution connected in the same app to an actual clinical delivery network."
The gap to watch: That integration is also a constraint. One Medical is US-only, limiting both reach and the diversity of real-world data it can learn from. Amazon’s decision to train on user interaction patterns has also attracted scrutiny from researchers who worry that health data could be shared too broadly inside a commercial ecosystem.
Perplexity: The data-unification play
Also this month, Perplexity launched Perplexity Health, on the argument that health data is structurally broken – scattered across portals, devices, and labs with no single view. Perplexity Health connects to Apple Health, major wearables, and electronic health records from over 2.4 million care providers. A Health Advisory Board of physicians pressure-tests decisions against evidence-based medicine standards.
📋 What Perplexity is betting on: Being the best answer engine for personal health – provider-agnostic, insurer-agnostic, system-agnostic. Maximum data breadth, minimum friction.
The gap to watch: Perplexity Health is rolling out in the US first. Building clinical trust without the institutional relationships OpenAI or Microsoft already have is the real test — and the Oxford study is a reminder that data connectivity alone does not guarantee better outcomes for real users.
Three companies taking a different path
Google is playing a longer game – building AI infrastructure underneath healthcare rather than launching a consumer product. At HIMSS 2026, Google Cloud announced enterprise partnerships with major US health systems using Gemini-powered AI agents for clinical workflows.
Meta is further back still, and deliberately so. Its open-source Llama models are being used by health startups and researchers across multiple continents to error-check radiology reports, cut clinical trial patient-matching from months to a single day, and power genomic analysis tools. Running Llama on their own infrastructure lets healthcare organisations keep patient data within their own walls – a meaningful advantage in regulated markets across the EU and UK.
Apple has not launched a health AI product yet. Apple Watch already detects sleep apnea, atrial fibrillation, and hypertension. The Health app is a de facto medical record hub for hundreds of millions of people, protected by on-device processing no cloud-based competitor can replicate. When Apple moves, it will do so carefully – and that caution may prove to be its strongest credential.
Limitations: what the headlines don't tell you
The promise is real but so are the gaps. Every product launched so far carries an explicit disclaimer that it is informational only and does not replace clinical care – and the evidence backs that caution up.
Real-world performance falls short of benchmarks. The Nature Medicine study showed that users interacting with leading AI models performed no better at assessing medical scenarios than people using standard internet search.
Access is deeply unequal. Every major consumer health AI launch of 2026 is, so far, US-only – excluded from the EU, UK, and most of the world by data privacy regulations. The people most likely to benefit are often those who can't yet access these tools.
Data privacy remains a live concern. Users are sharing their most sensitive personal information with large commercial platforms. Independent scrutiny of how that data is stored and used is still limited.
Liability is unresolved. When an AI health assistant gives incorrect guidance and a patient acts on it, who is responsible? None of the 2026 products has a clear answer.
A lot can change quickly in this space – and the pace of development suggests it will. For now, the gap between what these products promise and what the evidence shows is wide enough to warrant caution. The kind of healthy scepticism, in other words, that good healthcare has always required.
Innovation highlights
🫀 RNA rewires cardiac repair. After a heart attack, lost heart muscle doesn't grow back – a major reason survivors develop heart failure. Researchers have developed an RNA therapy that turns skeletal muscle into a temporary drug factory. A single injection into the arm produces a healing molecule that travels through the bloodstream and activates inside the heart, where a naturally occurring enzyme converts it into its active form. In preclinical studies, one injection reduced scar tissue and improved heart function for at least four weeks.
🩹 Wounds get a wireless watchdog. Scientists have developed a portable sensor chip that monitors four wound biomarkers at once – pH levels, uric acid, a bacterial compound, and an inflammation protein – to catch infection or inflammation before symptoms worsen. Built with laser-induced graphene, the chip transmits real-time data to a mobile app, with future smartphone integration planned. The team also aims to pair it with microneedle technology for continuous, pain-free monitoring of chronic wounds.
🧬 CRISPR catches every STI. Researchers built a rapid point-of-care (POC) test that detects all major bacterial STIs – including syphilis – from a single sample in under an hour. Using CRISPR-Cas enzymes programmed to recognize specific pathogens, the device screens for multiple infections simultaneously without lab equipment. It also flags antibiotic resistance in gonorrhea, helping doctors prescribe the right treatment faster. The goal: diagnosis and treatment plan in a single clinic visit.
Company to watch
🤖🧪 Insilico Medicine is a generative AI drug developer that has built one of the most advanced AI-powered discovery platforms in the industry. Founded by CEO Alex Zhavoronkov, the Hong Kong-listed company has developed at least 28 drugs using generative AI tools, with nearly half already at a clinical stage. Insilico develops its AI outside of China, in Canada and the Middle East, but conducts early preclinical drug development in China – and says AI can synthesize molecules more quickly than traditional methods.
The company's latest milestone is a global licensing and research deal with Eli Lilly, giving Lilly an exclusive license to develop, manufacture, and commercialize preclinical oral drug candidates across selected disease areas. The two companies have worked together since signing an AI-based software licensing agreement in 2023. Andrew Adams, Lilly's Group Vice President of Molecule Discovery, called Insilico's AI-enabled discovery "a powerful complement" to Lilly's clinical development.
Image source: Longevity.Technology
Weird and wonderful
🧊🧠 Freeze now, think later. Scientists in Germany have successfully rebooted mouse brain activity after deep-freezing brain tissue to –320 degrees Fahrenheit using liquid nitrogen.
The trick is vitrification – cooling tissue so fast that molecules get trapped in a glass-like state instead of forming ice crystals, which would otherwise rupture cellular walls. After spending up to seven days in the deep freeze, thawed brain slices showed intact neuronal membranes and, remarkably, preserved hippocampal long-term potentiation – the cellular backbone of learning and memory. The neurons even responded normally to electrical stimuli.
Lead neurologist Alexander German already has preliminary data showing the technique works on human cortical tissue, with ambitions to scale up to whole organs.
Image created using Canva AI
Thank you for reading the Healthy Innovations newsletter!
Keep an eye out for next week’s issue, where I will highlight the healthcare innovations you need to know about.
Have a great week!
Alison ✨
