The Interface That Became the Introduction

In 2009 a man in a city could open a map on his telephone and see who else nearby was interested. In 2012 a company reduced the whole decision to a gesture of the thumb. Within about fifteen years of those two events, the single most common way that two strangers in a wealthy country became a couple was neither a place nor a person. It was an interface, and the interface belonged to somebody.

That is a very large change to have happened quietly, and it is unusually well documented, which is the reason it belongs in this chamber rather than in the genre of complaint. We have nationally representative survey evidence from the United States, from Switzerland and from Japan, and we have the filings of the companies themselves. A reader does not have to take anybody’s impression of what dating has become.

The argument of this essay is that the interface is not neutral furniture. A product whose income arrives monthly from each person still using it, and a person whose stated goal is to stop needing the product, want different things. That divergence is structural. It is not a story about weak-willed users who should log off, and telling someone to be more disciplined is telling them to solve an asymmetry of design and information on their own time.

And the essay is going to refuse, hard, the conclusion that usually follows from that argument — that these things ruined intimacy. The same body of evidence shows that they materially widened who could meet whom, and for people standing in a thin market, including a great many queer people, that is not a small thing. Any account which ends by wishing the instrument away is asking a particular group of people to pay for everybody else’s nostalgia.

The counter-reading, which the essay takes seriously enough to build a section around, is this. The most common complaint about these services — the exhaustion, the sameness, the sense of being one of very many — is a description of abundance rather than of technology. Abundance is what was promised. Abundance is what was delivered. The difficulty may be that we asked for a larger room and discovered that a room with no walls is not a larger room.

The anchor is a paper by Michael Rosenfeld, Reuben Thomas and Sonia Hausen, published in the Proceedings of the National Academy of Sciences in 2019 and drawing on the How Couples Meet and Stay Together survey. Its title states the finding: disintermediating your friends. Among heterosexual couples in the United States who met in 2017, about thirty-nine in a hundred met online, against about twenty-two in a hundred in 2009. Meeting online overtook meeting through friends around 2013 and has not been overtaken since.

The shape of the change is the interesting part. Meeting through family, through the neighbourhood and in church had been falling since the middle of the twentieth century. Meeting through friends held up for much longer and then fell too, from the middle of the 1990s. What happened was not that people stopped being introduced. It is that the introducer changed from somebody who knew both people to a company that knows neither of them and holds a great deal of data about both.

The pattern is not confined to one country. Gina Potarca, working with the Swiss Federal Statistical Office’s 2018 survey of families and generations, found that around a quarter of relationships begun in the two preceding years had started online, and that apps had displaced websites as the dominant online venue after 2016. The Swiss case is useful precisely because Switzerland is not the United States: the same instrument, arriving in a different society, produced a recognisably similar curve.

Japan’s own series is now good enough to sit beside those. The National Institute of Population and Social Security Research published the results of its seventeenth fertility survey in September 2026, from fieldwork conducted in June 2025. Among couples in their first marriage who had married within the previous five years, the share who said they met through the internet — the questionnaire’s note specifies exchanges on social networking services, websites and apps — was 20.2 in a hundred, against 11.0 four years earlier. Meeting at work was 20.9, through friends or siblings 20.2, at school 14.0.

Read by year of marriage, the same survey is starker. Among couples marrying between 2015 and 2019, what Japan calls a love match accounted for 83.6 in a hundred and meeting through the internet for 6.9. Among those marrying between 2020 and 2024, the love match had fallen to 71.4 and the internet had risen to 18.6, overtaking the arranged introduction, which barely moved. Three countries, three instruments, one curve.

If the introducer is now a company, it is worth saying plainly which companies. Match Group tells its own shareholders that it owns and operates Tinder, Hinge, Match, Meetic, OkCupid, Plenty Of Fish, Pairs, Azar, BLK, Chispa, Upward, Salams, HER, Archer, Yuzu and The League, among others. For 2025 it reported revenue of about 3.5 billion United States dollars.

Bumble Incorporated, the other large listed operator, reported revenue of about 966 million dollars for the same year across Bumble and Badoo, having sold or discontinued several smaller apps. Grindr is separately listed and separately owned. Below those, a long tail of small applications competes for what is left. This is not a market of many sellers that happens to include a few big ones. It is a market with a very short head.

The concentration reaches into the categories one might expect to be exceptions. Archer, aimed at gay men, was launched by Match Group in 2023. HER, aimed at queer women, was acquired by Match Group in 2025. The apps built for people who were badly served by the mainstream have largely been bought by the mainstream, and the reason is not sinister: an audience that cannot easily meet offline is a reliable audience, and reliable audiences are acquired.

Japan is a partial exception and should be recorded as one. Pairs, the largest service, is operated by Eureka and sits inside Match Group’s Asian segment. But Tapple belongs to the CyberAgent group, and with and Omiai sit under a domestic holding company formed in the last few years. A reader in Tokyo is not simply looking at an American duopoly, and an essay that said otherwise would be describing the United States and calling it the world.

One episode says more about what an introduction has become than any amount of argument. Grindr was bought by a Chinese company and then sold again, under 2020 pressure from the United States committee that reviews foreign investment, on the reasoning that its holdings of location data, chat records and health information about millions of users constituted a national security exposure. Set the geopolitics aside and look at the premise both governments accepted without argument: the record of who you want is now an asset with a custodian.

Here is the structural point, and it can be made without attributing bad faith to anybody. Match Group reported that its revenue in 2025 was flat at about 3.5 billion dollars, that it ended the year with fewer people paying it than a year before — about 14.2 million — and that the money it took from each paying person went up. Fewer people, each paying more, producing the same total. That is a legitimate way to run a business. It is also a description of a business that does not require more people to succeed at the thing they came for.

The mismatch is narrower than the usual accusation and therefore harder to dismiss. Nobody needs to design for failure. A person who finds a partner leaves and stops paying; a person who is enjoying the search stays and keeps paying; the difference between those two outcomes, at the level of a quarterly result, is simply duration. A firm does not have to want you to fail. It only has to be indifferent between your finding someone in March and your finding someone in November, while you are not indifferent at all.

The companies know what the suspicion is. Match Group describes Hinge, in a filing to the United States securities regulator, as the dating app for people who want to get off of dating apps. That is a remarkable sentence to put in front of investors, and it is honest about the market it addresses: a product sold on the promise of its own obsolescence, to people who have already concluded that the category has a problem.

The public record contains harder material too, and it should be stated with its status attached. In August 2025 Match Group agreed to pay 14 million dollars and accepted an order to resolve charges brought by the United States Federal Trade Commission concerning a guarantee whose conditions were not adequately disclosed, the suspension of accounts belonging to people who disputed charges, and difficult cancellation. Match Group denied the allegations and admitted no liability. Separately, a group of users filed suit in California in 2024 alleging that several of the company’s apps were deliberately designed to be habit-forming; a court sent the claims to arbitration in 2025, so the allegations were never tested. The essay takes neither as proof of intent. It takes them as evidence that the question is live enough for a regulator and a court to have handled it.

What follows from all of this is modest and, we think, correct. The instrument is not a conspiracy. It is an ordinary business whose interests run parallel to the user’s for most of the journey and then diverge at the end, at exactly the moment the user cares most. Knowing that does not tell anyone what to do. It does tell them to stop reading their own fatigue as a personal deficiency.

Now the other half, and it is not a courtesy. In 2012, in the American Sociological Review, Rosenfeld and Thomas had already identified who benefits most from a searchable population: people facing what economists call a thin market, where the set of plausible partners in daily life is small. Gay men and lesbians were the clearest case. They were early adopters not because they were more technological but because the alternative was worse.

The 2019 paper puts a number on how that turned out. Among same-sex couples who met in 2017, roughly sixty-five in a hundred met online — a considerably larger share than among heterosexual couples, and an early signal of where everybody else was heading. Survey work by the Pew Research Center in the United States points the same way: adults who describe themselves as lesbian, gay or bisexual report having used these services at roughly twice the frequency of straight adults.

Sit with what that means for a particular person. Someone living in a town where the set of people who might plausibly want them is very small; someone whose desire is not legible in any room they are expected to be in; someone who cannot ask a colleague to introduce them without disclosing something they have not chosen to disclose. For that person the interface is not a degraded substitute for being introduced by a friend. It is the first introduction mechanism that ever existed.

The widening is not only about orientation. Potarca’s Swiss data found that couples formed through apps were more educationally mixed than couples formed elsewhere, an effect driven largely by women with university degrees partnering with men without them, and that they were drawn from a wider geographic area. In the same data, women who met their partner through an app expressed stronger intentions to move in together and stronger intentions to have a child within three years than women who met their partner offline. Whatever these services do, they do not only produce churn.

So the honest summary is uncomfortable for everybody. The instrument that produced the exhaustion is the same instrument that produced the introductions. There is no version of the argument in which one can be kept and the other returned.

Here is the counter-thesis, and we think it is the strongest case against this essay’s own framing. Almost everything people say against these services describes abundance, not technology. Too many options. Nobody choosing. Everybody comparing. The sense of being a line in somebody’s queue, and of running a queue of one’s own that never empties. None of that is a property of a screen. All of it is a property of a very large set of candidates arriving one after another with no natural end.

And abundance was the promise. It was not the fine print; it was the pitch. Rosenfeld himself put the advantage plainly in 2019: dating websites have enormous advantages of scale, and even if the greater part of the pool is not to your taste, a larger choice set makes it more likely that someone in it suits you. That is true. It is also the whole of the difficulty.

The psychology has been measured. Tila Pronk and Jaap Denissen, in Social Psychological and Personality Science in 2020, ran three studies of what they named a rejection mind-set: as participants worked through a sequence of potential partners, they became steadily more dismissive, with the chance of acceptance falling by about twenty-seven in a hundred between the first option shown and the last. The decline was explained by falling satisfaction with the photographs and by a falling sense of one’s own success. The effect is not a fact about the kind of person who uses an app. It is a fact about what a queue does to the twentieth judgement.

The same doubleness appears in what users report. In the Pew surveys, women who have used these services describe being overwhelmed by the number of approaches they receive, while men describe the opposite — the silence of receiving almost none. Those are usually treated as two complaints. They are one machine seen from its two ends, and neither end is having the experience the product’s photographs depict.

This Library has already made the adjacent argument, in its essay on the expectation gap and on what happens when choosing a partner becomes a strategy. That piece showed that one variable moved enormously while another did not move at all, and that the arithmetic, rather than anybody’s character, produced the distance between what people expect and what stands in front of them. We are not going to repeat it, and a reader who has not read it should.

What this essay adds sits one layer down. That essay was about the criteria. This one is about the queue. The criteria were in place before anyone had a smartphone; the instrument did not invent comparison, and it did not invent the strategy. What it did was remove the point at which comparison naturally stops. In a room, the set of people is finite and the evening ends. In a queue, the set is not finite and the evening does not end, and a person who would have chosen well among nine will not necessarily choose at all among nine hundred. That is the mechanism, and it is not anybody’s fault.

The usual way to write about Japan here would be to say that it arrived late and is catching up. The record does not support it. What Japan did was compress the whole social transition — from something you would not tell your mother about to something you would say at a wedding reception — into roughly a decade, and then do something no other country has done at the same scale.

The compression is visible in the numbers already given. Between the sixteenth and seventeenth fertility surveys, four years apart, the share of newly married couples who met through the internet went from 11.0 in a hundred to 20.2. A separate government instrument gives a larger figure again: a web survey run by the Children and Families Agency in July 2024, of twenty thousand people aged fifteen to thirty-nine, found that among those who had married in the previous five years, 25.1 in a hundred named a matching app as how they met, ahead of the workplace at 20.5. The two numbers are not the same, they were not produced the same way, and this essay does not average them. Both point in one direction.

The part with no equivalent elsewhere is the state. The Children and Families Agency runs a grant for regional measures against the falling birth rate, and among the items it supports at three-quarters of cost is the upgrading of matching systems, artificial intelligence named first. By the end of the 2023 financial year, thirty-two prefectures had introduced an AI matching system, and thirty-seven of the forty-seven prefectures were running a marriage support centre. In September 2024 the Tokyo Metropolitan Government launched an AI matching service of its own, open to single adults living, working or studying in the capital, requiring a certificate of single status and an interview, at eleven thousand yen for two years.

In 2025 the Digital Agency went further, agreeing with the industry association that the national identity card could be used to verify identity and marital status on private matching services, with Pairs and Tapple among the first. Consider what that is. Elsewhere the state’s posture toward these companies is consumer protection, competition law and safety regulation. In Japan the state is simultaneously a regulator, a subsidiser, a competitor and an identity vendor to its competitors.

And the reason given is not loneliness, and not safety, though safety is the reason given for the identity card. The programme sits inside the architecture of measures against the falling birth rate. That is a different framing from anywhere else, and it should be stated precisely rather than mocked: a Japanese citizen who opens a matching app is using an instrument that a portion of her government has an explicit demographic interest in her using well.

This Library does not read a birth rate as a scoreboard, and the reason is not squeamishness about demography. A birth rate is an aggregate of decisions each of which belonged to a particular person and was made inside a particular life. Reading the aggregate as a national performance indicator quietly converts those decisions into a duty owed, and the duty falls on the people whose bodies and years are involved, which is to say overwhelmingly on women.

The objection is not merely ours. In 2024 the external reviewers who examined this very grant programme for the Children and Families Agency wrote that behind the decline in marriage lie fundamental problems of income and employment that the programme cannot solve on its own, and that it therefore has to be run alongside wage increases and reform of the labour market. That is a government document saying, in its own voice, that the introduction is not the bottleneck.

We would put it slightly differently, and the difference matters. A state may perfectly legitimately want people who want partners to be able to find them, and may spend money on it, and may run it better than a company optimised for monthly revenue would. That is a service, and a defensible one. What changes the character of the thing is the measure of success. If the measure is whether the person is glad they met, the service has one customer. If the measure is the number of children, the service has acquired a second customer, and the second customer is not the person in the room.

There is a further group the apparatus tends to render invisible. A person who does not want a partner, or does not want children, or wants both but not yet, is not a shortfall in a national series. This Library’s position on that has been consistent and is not going to soften here: they owe nobody an account of themselves, least of all a statistical one.

This house sells an introduction that is not an app, and it profits from app fatigue. That should be the first sentence of this section rather than the last, because every paragraph above — about queues, about revenue per paying user, about the rejection mind-set, about the twentieth judgement being worse than the first — describes a condition for which this house happens to be the alternative. A reader should discount the essay accordingly, and we would rather hand them the discount than have them find it.

What the house can honestly say about the difference is narrow. An evening arranged here is a finite set with an ending built into it, put together by a person who is accountable for having put it together, and the person on the other side of it is not selecting from a queue while they are in the room. That is a different structure from an infinite feed, and the difference is not marketing. It is also, plainly, a service with a price, available to people who can pay it, which an app is not in the same way.

What it may not claim is longer. It is not a route to a partner and must never be sold as one. It does not scale, and nothing in this essay should be read as a proposal for how a society should arrange its introductions. It is not a criticism of anybody’s use of an app, and a person who met their partner in a queue did not do it wrong — a great many people did exactly that, including a fifth of everyone who married in Japan in the last five years. And nothing here is advice to delete anything.

The most a bounded evening can honestly offer, against the material above, is one hour that is not an audition: an interaction in which nobody is comparing, because there is nobody to compare against, and in which the ending was agreed before the beginning. Whether that changes how the queue feels on the following Tuesday is not in this house’s gift, and we do not claim it.

It does not claim that these services ruined intimacy. That conclusion is available from half the evidence and is contradicted by the other half, and the essay has tried to hold both halves at once rather than choosing the more quotable one.

It does not claim that anybody designed an addiction. The 2024 litigation alleging deliberately habit-forming design was sent to arbitration without its allegations being tested, and it is cited here for its existence and nothing further. The 2025 settlement with the Federal Trade Commission concerned advertising, billing and cancellation, was resolved without any admission of liability, and is not evidence about how matching works.

It does not claim that the Japanese, American and Swiss figures are comparable. They come from different instruments with different questions, different populations and different definitions of meeting online, and the Japanese meeting-circumstances table rests on a modest number of couples — five hundred and seventy in the latest wave, seven hundred and eighteen in the one before — with the internet option available for only two waves. The direction is shared. The magnitudes are not interchangeable.

It does not claim that ownership is equally concentrated everywhere. Japan’s market has substantial domestic operators, and the essay says so rather than exporting an American market structure to the world.

It does not claim that the instrument caused the fall in marriage, the fall in partnering or the fall in births. The surveys it relies on interview couples who formed, which means they cannot see the people who stayed in the queue, and a method that only ever meets the winners cannot settle what happened to everybody else. The corporate figures are the companies’ own, filed for investors rather than for sociologists. The widening of who could meet whom is measured most clearly in the United States and Switzerland, and the essay does not assume it transfers unchanged. And this piece was written from outside every market it describes, about an instrument that changes faster than the surveys that measure it.