.svg)
It's time America took matchmaking seriously.
In 2023, South Korea's total fertility rate hit 0.72. The lowest in recorded human history. Every op-ed about demographic collapse for the next eighteen months used it as the headstone.
Then the number went up. In 2025 it hit 0.80. Still catastrophically below replacement, but moving in the right direction for the first time in nearly a decade. The rebound was not mysterious. Marriages rose 14.8% in 2024 and another 8.1% in 2025. Births followed, with the usual one-to-two year lag.1
In the United States, the numbers are moving the other way.
Only 60% of 35-year-old American men have ever been married, down from 90% in 1980. The US fertility rate hit a record low in 2025, falling below 1.6. Replacement requires 2.1. On current trends, closer to half of today's young adults will never marry.2
Fertility is only the most visible consequence of what's broken. Marriage produces outcomes across nearly every variable a functioning country cares about. Children raised by married parents have roughly a quarter of the poverty rate of children in single-parent families (8% versus 35%), and this gap persists even after controlling for family income. Chicago Fed research attributes much of the decline in US social mobility to declining marriage rates among low-income Americans. Married adults live longer, report better health, and survive serious illness at higher rates than never-married or divorced peers. Married people consistently report higher life satisfaction. Communities with higher marriage rates have more civic participation and more social trust.3
Stack these on top of each other and marriage becomes more than a lifestyle preference. It's load-bearing infrastructure for most of the things Americans say they want from their country.
So what's breaking it? Demographer Lyman Stone has the cleanest decomposition of the question. Depending on which survey you look at, around 80-90% of Americans still want to marry, and desire for children has remained broadly stable in recent decades. Married fertility has been essentially stable too. Roughly 75% of the US fertility decline since 2007 is attributable to people not marrying in the first place, not to married couples choosing to have fewer kids. Americans still want to get married and have children. The question is why they have stopped doing it.4
The answer is infrastructural.
The usual candidate explanations are real and matter. Housing is expensive. Wages have stagnated for men without college degrees. Norms around family formation have shifted. Young adults reach the traditional milestones of adulthood later than previous generations did.
My argument is narrower: on top of all of that, America also dismantled the intermediary layer that used to produce marriages, and the software that replaced it is optimized to keep people from forming them. This is the piece nobody is building for.
For most of human history, marriage was produced by an institution found in every documented civilization: formal matchmaking. Edward Westermarck's 1891 survey The History of Human Marriage went through every culture he could document and found designated matchmaking roles in all of them.5 The shadchan in Jewish communities. The nakodo in Japan. Village go-betweens in China. Matrons in Rome. Priests, relatives, professionals.
These roles operated within three overlapping social structures. Extended family. The local community, usually anchored by a religious institution. And the geographic village. Together they introduced eligible people to each other. They vetted. They applied social pressure toward commitment. They caught drift early.
The idea that an American identifies, evaluates, and commits to a spouse entirely on their own is a sixty-year experiment, not the historical human baseline.
Extended family is scattered across time zones. Religious attendance has collapsed. The geographic village does not exist for most educated Americans, who move for school, move for first job, move for second job, and rebuild a social graph from scratch each time.
Dating apps replaced introductions at scale, but not the vetting, the accountability, or the commitment pressure. They are entertainment products optimized for engagement, not pair formation.
An app optimized for engagement is, by definition, optimized against exit. The design incentive of every major dating app is for you to remain a user. Marriage makes you stop being a user. The product is therefore structurally misaligned with the outcome its customers actually want.
The symptoms of this misalignment are easy to see. The category ships unisex product experiences for what is, statistically, one of the most gender-differentiated behaviors humans engage in. The one major player to deviate from the unisex template was Bumble, which built its entire product around women messaging first. In every documented human context across cultures and centuries, men initiate courtship at far higher rates than women. It is one of the most cross-culturally stable behavioral asymmetries humans have.
Bumble took that asymmetry, inverted it, and shipped the inversion as the core product mechanic. The result is what you would expect when you design against the behavior you're trying to facilitate: women pushed into an initiator role most do not naturally take, low-effort opening messages, mass ghosting, and engagement without outcomes. It was cosmetic feminism as a product strategy, and industry coverage celebrated it as innovation.
A dating product's most basic filter is the sex of the people its users want to date. In 2016, Tinder quietly overrode that filter by default, announcing in a press release that transgender women would now appear in heterosexual men's match queues, and called it inclusion.6 The design choice is hard to defend on product terms. A product that exists to match people based on mutual attraction deciding that mutual attraction is a form of discrimination to be engineered around is not a serious marital infrastructure company. It is an ideology wearing the skin of one.
Hinge markets itself with the slogan "Designed to be deleted." It is one of the most polished brand campaigns in consumer tech.7 The product itself uses daily like caps, subscription paywalls for unlimited swipes, and algorithmic pacing designed to keep users engaged long enough to pay. If the product were actually designed to be deleted, it would not be monetized on a subscription funnel. The slogan is a mission statement and a user-acquisition claim at the same time. The underlying product refutes both.
Look at who runs these companies. Bernard Kim, Match Group's CEO from 2022 to 2025, spent his career as a mobile games executive — nearly a decade at Electronic Arts and six years as president of Zynga. Profiles of his tenure described him "leveling up" dating apps using the monetization playbook he built for mobile games. His successor, Spencer Rascoff, met his wife at age 17, before Tinder, eHarmony, or Match.com existed. He has been married for more than thirty years — longer than any modern dating app has been in operation. By his own telling in a recent Business Insider profile, he has never dated through the products he runs.
This is not a knowledge problem. Match Group employed Dr. Helen Fisher, one of the most prominent biological anthropologists of her generation, as Chief Scientific Advisor for nearly two decades. She spent her career studying the neurobiology of romantic attachment and mate selection. With her research available the entire time, the products still came out looking like Tinder and Hinge. The expertise was on the payroll. The business model wouldn't let it touch the product.10
Users have noticed. Every major dating app has a negative Net Promoter Score. More users would actively recommend against these products than for them. Tinder's is −52%. This is where modern Americans go to try to find a spouse.8
The courts have noticed too. In February 2024, a class of Tinder, Hinge, and The League users — all Match Group properties — sued Match Group in federal court alleging that the apps are intentionally designed to be addictive, compare their own internal mechanics to gambling, and violate consumer protection laws. The complaint describes Match's business model as dependent on "fomenting dating app addiction that drives expensive subscriptions and perpetual use" and accuses the company of designing features "to gamify the platforms to transform users into gamblers locked in a search for psychological rewards that Match makes elusive on purpose." Match denied the allegations. Whether or not the suit succeeds legally, the fact that the complaint describes the actual product accurately is the problem.9
This is not a product category having a rough patch. It is category-wide infrastructure failure, operating on the most important decision most people make in their lives. At that scale, the problem stops being a business opportunity and starts being a public failure — the kind that deserves the seriousness we bring to rebuilding an electrical grid or a semiconductor supply chain.
Marital infrastructure gets built toward a different goal. Measured on marriages produced, not time-in-app. Unit economics only make sense if the success rate is high enough to justify the acquisition cost, because each user churns out of the product exactly once.
At the product level, marital infrastructure inverts the dating-app defaults:
Each inversion is forced by the outcome metric, and none of them are compatible with a swipe-based subscription product.
I know the shape of this works because I am building it. At Keeper, roughly 10% of our first dates end in marriage. That is orders of magnitude above any other matching system at scale. The ratio describes a different product category.
The category has a surface-resemblance problem. To investors, anything with a consumer interface reads as "consumer app." Swipe mechanic. Subscription funnel. Optimize for DAU. Founders misdiagnose the category the same way. A wave of new startups has emerged pitching the idea that what's missing is a quirky mechanic — it's all SMS-based, or it's a voice agent, or it runs inside a chatbot. These are surface changes dressed as product innovation. None of them address where the difficulty actually lives.
The hard part of marriage matching is the underlying science of who matches with whom and why those matches last. Solving that requires proprietary data, models, and iteration. No input modality substitutes for that work. Not swipes, not SMS, not voice, not video.
Solving it also requires following the science wherever it leads, including into the parts of human mating that are politically inconvenient to discuss. Keeper works with evolutionary psychologists and relationship scientists. They take the evidence seriously regardless of whether the findings fit current sensibilities. A product that flinches from the science cannot build the models, because the models are the science.
Interface still matters, and the quirky-mechanic startups aren't wrong that simpler is better. Less friction keeps people engaged. But there's a floor on how simple you can go. Real compatibility signal lives across many surfaces: video, preference rankings, appearance rating, psychometric tests, behavioral data, digital footprints. No single mechanic captures them all. The startups racing toward hyper-minimal are building as if that floor doesn't exist.
Keeper runs proprietary models that outperform frontier LLMs on compatibility at a fraction of the cost. The interface is built to collect those signals. The R&D stack is built to interpret them.
The category is empty because the problem is hard, and the obvious path (another dating app) produces marriages as a byproduct of engagement, not as the north star metric. The non-obvious path requires R&D depth that screens out most founders willing to try.
A common objection I hear is that matchmaking fails due to adverse selection. The premise is that high-status, high-agency people do not need matchmaking because they can find partners on their own, so the pool must be populated by the people who couldn't. As if they're defective. Some users do fit that profile. But the inefficiency of spouse search scales with opportunity cost. The most ambitious people have the least time to waste on it, and the most specific requirements when they do. They are the most frustrated by the current system, not the least. Keeper's signups skew heavily toward that cohort. The 10% number is driven by them.
They are also, not coincidentally, the cohort that staffs everything Americans are now trying to build. Every rocket, every reactor, every lab, every production line is built by people who exist only because their parents met. Marriage is the input to everything else.
If Von Neumann's parents had been born in 2000, they would be trying to find each other on Hinge. They would have never met.
Picture America in 2050 on the current trajectory. Closer to half of adults reaching middle age having never married. Household formation continuing to collapse. Fewer kids in classrooms, fewer first-time buyers, fewer young people starting businesses, a shrinking working-age population taxed ever more heavily to support a growing retired one. Less trust, less civic participation, fewer Von Neumanns born. All of it downstream of a single broken pipeline.
We are spending hundreds of billions on defense technology to protect this country. The total venture capital invested in the entire dating category over the past decade is a rounding error by comparison, and almost all of it went to engagement products that make the problem worse. If we don't invest in marital infrastructure now, we will look back in thirty years and wonder what we were trying to protect.
This is an open research problem. Academic psychology and sociology have studied mate preference and marital success for decades, but the experiments that would actually establish what single-shot matchmaking requires have never been run. Running real matches at scale. Measuring outcomes over the years and decades that matter. Iterating on the match function with real ground truth. The data does not exist because no one has been funded to produce it. There is no fundamental reason it can't be figured out. It's the problem a serious country would treat as a top priority.
It also cannot be figured out by iterating on the existing dating app flywheel. Any product that asks users to keep engaging, keep swiping, keep returning to the app will inevitably collapse toward whatever use case generates the most engagement. In dating, that use case is not marriage. It is casual dating, validation, and hookups, because those are the behaviors that keep users in the funnel. A serious marital product has to be designed for exit, from day one. Matches made once, not harvested continuously. Behavioral controls that protect the outcome from being cannibalized by the funnel. The existing category cannot pivot into this product. Any serious attempt would require Match Group or Bumble to voluntarily collapse their subscription revenue while proving out a new product category, a multi-year revenue trough that public-company incumbents structurally cannot survive. The economics of its business model do not permit it.
The question is whether we treat this as a serious civilizational research problem or whether we continue outsourcing it to companies optimizing DAUs and retention curves.
The dynamism of America is not possible without marital infrastructure. It is time America took matchmaking seriously.
Build the infrastructure.
1 South Korean TFR and marriage data from Statistics Korea, February 2026 release. The 2023 TFR of 0.72 was the lowest ever recorded globally; 2025 TFR reached 0.80 after marriages rose 14.8% in 2024 and 8.1% in 2025. Coverage: Reuters, Nikkei Asia.
2 US fertility data from CDC preliminary 2025 natality data (general fertility rate 53.1 per 1,000 women 15-44, down from 53.8 in 2024; TFR below 1.6). US marriage data from Lyman Stone, Institute for Family Studies: "1-in-3: A Record Share of Young Adults Will Never Marry" for the 35-year-old men ever-married figure and the 1-in-3 lifetime projection, which is Stone's published conservative estimate; trend data since (see chart) suggests the true number for today's young cohort is closer to half.
3 Child poverty rate comparison (8% in married-couple families vs ~35% in single-parent families) from US Census Bureau Current Population Survey data, with academic analysis in Manning & Brown (2006) and related peer-reviewed literature. Social mobility decline tied to marriage decline: Davis & Mazumder, Federal Reserve Bank of Chicago. Health and longevity benefits of marriage: Lawrence et al., Journal of Happiness Studies 2019; Harvard Medical School findings on cancer survival and cardiovascular outcomes. Child outcomes beyond poverty (abuse rates, substance use, school completion) per Manning & Brown and related literature.
4 Lyman Stone, Institute for Family Studies Pronatalism Initiative. The ~75% decomposition of US fertility decline since 2007 is from Stone's "Marriage Still Matters" research and related IFS publications. Stated-preference figures for marriage fall in the 80-90% range across surveys; Stone's ~90% figure comes from his September 2024 podcast appearance "Why Is Everyone Having Fewer Children?" (timestamp 53:28). For related work, see Stone on Macro Musings (Mercatus Center) and IFS publications at ifstudies.org.
5 Edward Westermarck, The History of Human Marriage (London: Macmillan, 1891). Free on Project Gutenberg. Westermarck's comparative survey drew on British Museum archives and questionnaires sent to observers living among cultures worldwide. The three-volume 1921 edition is the expanded reference.
6 Tinder's "All Types All Swipes" campaign launched November 2016. Coverage: TIME, Engadget.
7 Hinge's "Designed to be Deleted" platform launched 2019 and has run continuously since. Creative Review coverage; Hinge CMO interview in Marketing Dive. Subscription mechanics and like caps are documented in the company's own product literature and the class-action complaint cited below.
8 MeasuringU 2024 online dating UX benchmark. All seven dating apps and all five websites studied received negative NPS; Tinder's was −52%, Match's −13%.
9 Oksayan v. Match Group Inc., 3:24-cv-00888 (N.D. Cal. filed Feb. 14, 2024). Quoted language from the complaint. Coverage: NPR, Washington Post, Reuters.
10 Bernard Kim served as Match Group CEO from May 2022 to February 2025. Previously President of Zynga (2016-2022) and SVP of Mobile Publishing at Electronic Arts for nearly a decade. Match Group profile; WSJ profile via Dallas Innovates on his "leveling up" approach to dating-app monetization. Spencer Rascoff succeeded Kim as Match Group CEO in February 2025 (GeekWire). The details on his marriage of more than thirty years and never having dated through the products he runs are from his March 2026 Business Insider profile, republished widely. Helen Fisher (1945-2024) served as Chief Scientific Advisor to Match.com from 2005 until her death; biological anthropologist, Rutgers / Kinsey Institute, author of Why We Love and Anatomy of Love, foundational MRI studies on romantic attachment neurochemistry.
