Health data became travel-insurance data that quarter.
In Q1 2024 a half-dozen AI travel-insurance platforms quietly shipped to market and almost no one in the trade press named the structural shift. The product looked the same: a checkbox on the booking flow, a $20-to-$80 add-on, a policy that pays out for trip cancellation, medical evacuation, the usual. What changed was the underwriting layer underneath.
The new layer takes the booking metadata, the destination metadata, the traveler's loyalty history, and (when the customer opts in, which most do because the discount is real) the traveler's wearable and pharmacy data, runs all of it through a model trained on five years of claims data, and prices the policy individually. Time from quote to bind is under a second. Time from claim filing to claim approval is under an hour for the standard claims and under twelve hours for the unusual ones. The payout is faster than the traditional broker. The price is lower for the median traveler. The marketing writes itself.
The marketing missed the structural mechanism. The structural mechanism is adverse-selection reconfiguration, and adverse-selection reconfiguration is the polite name for breaking a particular kind of insurance market.
A traditional travel-insurance pool works because the actuarial averaging happens across a population that the insurer cannot fully discriminate within. The 75-year-old cardiac patient and the 28-year-old marathoner pay roughly similar rates because the insurer cannot, in practice, price them individually at the booking moment, and the regulation in most jurisdictions limited what they could ask. The pool subsidized the high-risk traveler at the expense of the low-risk one, and the comprehensive product existed because both sides showed up. The 28-year-old got a lower rate than they would have gotten on a fully-priced individual policy. The 75-year-old got a policy at all.
AI underwriting at the booking moment changes both halves of that math.
It does not just price more accurately. It prices the high-risk traveler out of the comprehensive product, because the model identifies them at booking, names a price they will not pay, and routes them to a more expensive specialty policy or to no policy. It prices the low-risk traveler closer to their actual risk, which is much lower than the pool average. The pool fragments. What used to be one product becomes a tier with the low-risk traveler at the bottom paying half what they used to and a tier at the top with prices that nobody buys.
The operator-side story is who keeps the data.
Booking platforms (Expedia, Booking, Amex Travel, the major OTAs) collect the data at the point of quote-to-bind. They keep it. The model improves on the platform's data, not the underwriter's. The underwriter, in the new model, is a paper layer with a balance sheet that absorbs the pool risk; the moat is the data graph the platform owns. The same shape that played out in payments — issuer becomes commodity, processor with the data wins — is playing out in travel insurance, on a 24-month curve.
That has implications the trade press is not yet pricing.
First, the legacy travel-insurance brokers are facing a margin compression they cannot underwrite their way out of. Allianz, AIG, and the regional brands have the actuarial expertise and the regulatory licenses, and exactly none of the proprietary data on the population that books trips. They will, of course, partner with the booking platforms. The partnership terms will, of course, transfer the data graph to whichever booking platform the partnership exits to.
Second, the medical-evacuation product becomes a different category. Medical evacuation has always been the part of travel insurance that travelers actually cared about; it was also the part that the actuarial pool subsidized most heavily, because the high-risk-of-medevac population was a small fraction of bookings. AI underwriting prices medevac risk individually. The 75-year-old cardiac patient who used to get medevac coverage rolled into their $50 trip-protection policy now gets a separate quote at $300, and most of them stop buying it, and the medevac providers (whose unit economics depended on the comprehensive-pool subsidy) face a demand cliff at the same time the AI underwriting platform reports record-low loss ratios and high satisfaction scores. Both numbers are true. They are about different cohorts.
Third, and the most important for an operator looking five years out, this is the canonical adverse-selection-reconfiguration shape. Health insurance is going to play out the same way. Auto insurance has already played out partway. Life insurance will play out next. The general pattern is: a model that prices individual risk accurately at the moment of binding fragments the pool in a way that benefits the median customer, harms the marginal-high-risk customer, and transfers the data moat to whichever platform sits at the binding moment. Travel insurance is the early case study because the regulatory environment is the lightest and the booking-moment data is the easiest to collect. The operator who internalizes the pattern in travel can apply it to the categories where the regulation is heavier and the optics are worse.
The thing that crosses pillars is what the title is naming. Health data became travel-insurance data that quarter, not metaphorically: the pharmacy refill record, the wearable resting-heart-rate, the prescription-fill-rate over a year are now load-bearing inputs into a $40 add-on at the airline-booking flow. The traveler did not opt into a health-insurance underwriting decision; they opted into a discount on travel insurance. The discount was real. The data graph the booking platform now holds is the same data graph that, three years from now, will be load-bearing for a category where the regulation is much heavier. The booking platform, of course, is in no hurry to clarify the second use to the customer.
The honest forecast is that this product category lands as a quiet success in 2024-2025, the trade press writes about it as an AI-improves-customer-experience story, the actuarial pool fragments quietly across the same window, and by 2027 the legacy travel-insurance brokers are either acquired-by or partnered-with the booking platforms or out of the category. The data graph the platforms accumulated in those four years is then deployed against the next adjacent category. The line between travel insurance and health insurance, on this curve, gets thinner not because the categories merge but because the same operator owns the data graph that prices both.
durable read: if you are an OTA in 2024 and you do not have an AI underwriting layer in your booking flow by Q3, you are leaving margin on the table that the next twelve months will price at a permanent disadvantage. If you are a traditional travel-insurance broker in 2024, your strategic decision in the next eighteen months is which booking platform you partner with, because the partnership terms determine whether you exist in 2027. If you are a regulator, the structural shift is the one to name. The product is faster and cheaper. The pool is fragmenting. Both are true.
—TJ