She Is Running Two Searches, and Only One of Them Knows How to Stop.

It is twenty to midnight and she has two tabs open. In the first, a dropdown is asking her what she is looking for, and the list is longer than the one she grew up with: a life partner, a long-term relationship, a long-term relationship but open to something shorter, something short-term but open to it becoming more, something short-term, still working it out. In the second tab a marriage agency is listing what to bring to the first interview — identification, proof of income, and a certificate from her ward office confirming that she is legally free to marry.

Same woman. Same week. The word the culture reaches for here is confused, and it reaches for it faster when she is thirty-four than when she was twenty-four.

She is not confused. She is running two searches, and they are not the same kind of search: one of them has a rule for when it stops, and the other has no such rule and was never built to have one. Almost everything said about modern dating — including most of what gets said about her — comes apart once that is on the table.

This essay is about what those two machines actually are, why the menu of intentions changed before anybody’s behaviour did, why the exhaustion women report is not loneliness, and what the measured literature says about who stays and why. It is written for the woman with both tabs open, who has been told she is confused by people who never once asked her what she was doing.

Start with what each one is selling, because they are not competitors and it clears up a great deal to see that.

A marriage agency sells three things: filtering, verification and an ending. The documents are the verification — a person who has produced a certificate of single status from a municipal office and a statement of income has been checked in a way nobody on an app has been checked. The profile fields are the filter. And the staged courtship is the ending: introduction, then a provisional stage in which seeing several people at once is expected and normal, and then a further stage which you enter precisely by closing the others. The process names the moment you stop looking and makes stopping a condition of going forward. That is not a side effect of the model. That is the product.

An app sells reach. Reach without verification, at a scale no agency can match, for a price that rounds to nothing — and with no stage anywhere in it that requires anything to end. Its architecture has no concept of finished. This is not a moral failing of the companies; it is what a shelf is. A shelf that empties is a shelf with a problem.

So: a funnel with a gate, and a shelf that never empties. The woman with both tabs open has not contradicted herself. She has noticed that one of these answers a question the other cannot, and she is using each for what it is good at — which is what a competent person does with two different tools.

What nobody sold her, in either tab, is the thing that sits underneath both: a way of deciding when enough is enough. The agency imposes one from outside. The app leaves it entirely to her. Most of the difficulty in this whole subject lives in that gap.

Five years ago the sayable answers on a Japanese profile were short and they all pointed one way: serious relationship, marriage in view, no games. The menu now offered by the major apps is materially different — one of them asks directly for dating intentions and lists six, including long-term but open to short and short-term but open to long. Where the options used to be a wall, they are now a spectrum with names on it.

Here is the part that requires discipline, because it is where commentary usually goes wrong. A rise in the number of women tagging themselves short-term is, in the first instance, evidence about the existence of the tag. It is not a measurement of conduct. We have not watched anybody’s behaviour; we have watched a form change, and we cannot read off a form how many people are describing something new about themselves versus describing accurately, for the first time, something that was already true.

That second possibility deserves more weight than it gets. Before the menu existed, a woman who wanted company, attention, weekends, an end to a long silence — and not, at that moment, a wedding — had exactly two options on a form: say marriage and mean something else, or say nothing and be read as saying marriage. The menu did not invent her. It gave her the first accurate box.

And it came with a cost, which is worth naming in the same breath. A category that can be chosen is also a category that can be assumed, and the same tag does not read the same way on every profile. A man who marks short-term is read as knowing his own mind. A woman who marks the same box is, in a great deal of the commentary about this shift, read as something else entirely. The asymmetry is not in the software.

This is the cruel part, and the anger belongs to the mechanism rather than to anyone in particular: the honesty the new menu made possible is priced differently depending on who is honest.

The tiredness women describe about apps is real and it is consistently misdescribed. It is generally reported as loneliness, or as disappointment, or as evidence of a thinning market. It reads far more like something else.

There is a body of work in decision science on what happens when a person searches an option set with no natural end. A satisficer sets a standard and takes the first thing that clears it. A maximiser tries to establish that nothing better is available — which, in an option set that refreshes every morning, is a task with no completion condition. In one of the better-known studies of the distinction, maximisers ended up with objectively better outcomes and felt measurably worse about them. Doing better and feeling worse is not a paradox here; it is arithmetic. Every additional option raises the cost of concluding.

The most careful review of online dating from psychological science made a related point about the browsing itself: working through profile after profile encourages an assessment mindset, in which people are held against criteria rather than met. The same review found that the matching claims the industry makes were not supported by the evidence. Both halves matter. The machine is very good at reach, is not doing what it says it does when it claims to know who suits you, and quietly trains the user into evaluation mode.

So the exhaustion is not the exhaustion of being alone. It is the exhaustion of an assessment that never switches off — running, in the background, through a Tuesday meeting and a Sunday lunch, because a search with no stopping rule stops being a project and becomes a condition you live inside.

Seen from there, what the agency sells is suddenly legible, including to people who find the whole apparatus unromantic. It sells the stopping rule. The documents, the stages, the counsellor’s phone call, the requirement to close the others before you can go further: an external structure that ends a search which the searcher, in an endless option set, cannot reliably end on her own.

The other half of this is what men are doing, and it deserves the same discipline as the first half, which means describing a structure rather than indicting a category.

Option visibility works on everybody. The same menu that let her say the accurate thing lets him say it too, and the same endless shelf trains him into the same assessment mode. What is different is the surrounding market: what a man can buy, summon or scroll in the gap between relationships has never been cheaper, more private or more available, and a great deal of what is on offer is explicitly designed to substitute for the parts of connection that are inconvenient to earn. No causal claim is being made about anybody’s conduct. The observation is about supply.

And then there are the clocks, which are the real asymmetry and the one that generates almost all of the mismatch everybody has noticed. The marriage market in this country prices a woman’s age explicitly — the search filters at an agency say so in numbers, which is at least honest — while pricing a man’s age barely at all. Biology contributes a genuine deadline on one side and a much softer one on the other. The social clock adds a second layer on top of the biological one and aims it in the same direction.

A market where one side faces a priced deadline and the other faces none will produce exactly the pattern everybody complains about: one side moving to close, the other side keeping the shelf open, and both sides experiencing the other as having changed. Neither has to have become worse for this to happen. The incentives are doing it.

That is worth anger, and the anger goes to the arrangement rather than to the men in it, most of whom are also improvising without a script. The specifications have shifted over the decades — the three highs of the bubble years gave way to a stated preference for the three averages — but the underlying structure, in which one party’s time is a countdown and the other party’s is an asset, has not shifted at all.

There is a sharp intuition waiting at the end of all this: that if everybody keeps their options open, then everybody is permanently one step from changing boats, and a society arranged that way runs on a low background hum of mistrust.

Half of that is a measured finding and the other half is the part nobody sells. In the most-used model of commitment in relationship science, commitment is predicted by three things: satisfaction with the relationship, the quality of the alternatives, and the size of what has been invested in it. Quality of alternatives is an actual variable in an actual model. The intuition is correct, and it has a name.

But investment is the third term, and it is doing more work than the first two. Investment is the shared history, the flat, the friends who came as a pair, the language two people built that works nowhere else, the things that would be lost rather than merely ended. A relationship with high satisfaction and low investment is more fragile than it feels. A relationship with high investment survives long stretches of mediocre satisfaction, sometimes for better and sometimes for very much worse.

What follows is not romantic and is genuinely useful: the variable you can act on is the one the entire market ignores. Nothing in this economy sells investment. Apps sell alternatives, agencies sell filtered alternatives, and every service in the middle sells satisfaction in an hour. Investment is slow, unphotographable, and impossible to price, so nobody advertises it — and it is the term that decides whether anybody stays.

It also cuts the other way, and this house is not going to pretend otherwise. High investment is why leaving is hard when leaving is right. The model does not say that staying is the good outcome. It says what predicts staying, which is a different sentence and a more honest one.

Step back far enough and the apps are not the cause of anything. They are the delivery mechanism for a change that was already most of the way through.

For most of the last century marriage was not primarily a choice. It was the default: the thing that happened unless something prevented it, at roughly the expected age, with the search ending on arrival. Defaults perform work that choices cannot. A default ends the search, and it does so without anybody having to justify the ending.

Sociology has been describing the replacement since the early nineties — a relationship continued only for as long as both parties judge it to be delivering, which is a description of an arrangement with no floor under it. The consequence is not that people love less. It is that staying now requires continuous justification, by both parties, forever, and that this is a new labour nobody was trained for and nobody counts.

In this country the numbers are unusually legible. Marriages fell below half a million in 2023, which the vital statistics had not recorded in about ninety years. The census measure of the proportion never married at fifty has climbed for decades and now stands at close to three men in ten and not far off one woman in five. The state has begun subsidising prefectural matchmaking programmes, including ones that match algorithmically. When a national government is paying to run a search service, the default is not weakening. It is gone.

None of that is decline, and the essay is not going to sell it as one. It is a transfer. The work of ending a search used to sit with an institution, a family, a neighbourhood and a workplace, all of which did it badly and coercively and often to a woman’s cost. It now sits with two individuals, who do it more honestly and with no instruction whatsoever.

The last strand is the one that makes the whole picture cohere: women deciding to become mothers without a partner, through a clinic, a donor or a private arrangement.

Marriage in its twentieth-century form was a bundle. Sex, companionship, children, money, housing, social standing and care in old age arrived in one contract, and the contract was the only way to get most of them. Bundles hold as long as the components cannot be bought separately. Every component in that list is now separately obtainable to some degree — unevenly, expensively, and far more easily for some women than others — and a woman who wants a child and does not want a husband is not doing something exotic. She is declining a bundle for the one item in it she actually wants.

Here that meets a wall, and the wall is institutional rather than moral. Donor conception has been practised in Japan since the late nineteen-forties, but the professional framework around it has been built for married couples, with a spouse’s consent as a structural assumption. The law passed in 2020 settled who counts as the mother and who counts as the father when a married couple conceives with donated material, and explicitly left the harder questions — records, anonymity, a child’s eventual right to know — to legislation that was still being argued about afterwards. A single woman is therefore largely outside the frame, which pushes her towards private arrangements or a clinic in another country, and that is precisely where the risk concentrates: medical risk, legal ambiguity about parentage, and a child who may one day have questions that nobody kept the records to answer.

That is a matter for a clinic and a lawyer, by name and without hedging: if this is your question, take it to a reproductive medicine specialist and to a family lawyer in the jurisdiction you would actually use, because the answers have moved recently and may move again.

And if you are the woman with the two tabs open: you are not indecisive, and you are not doing anything incoherent. You are the first generation for whom this arrangement is genuinely a choice rather than a schedule, and nobody has written the manual because nobody has needed one before. The tiredness you feel at midnight is not a character flaw and it is not evidence that you are doing it wrong. It is the cost of running an open-ended search without a rule for when to stop — and deciding what your rule is, before the shelf decides for you, is the one piece of this that is entirely yours.

It offers no figure for how many women describe themselves as looking for anything in particular, because the honest answer is that a profile tag is evidence about a form and not a measurement of conduct, and there is no reliable public count of the shift itself.

It does not claim that marriage is finished. Most unmarried people in this country continue to say in national surveys that they intend to marry, and the essay’s argument is about the disappearance of the default rather than the disappearance of the institution. Those are different claims and only the first one is being made.

It takes no position on which search anybody should run, and it assesses no named service. The description of how a marriage agency works is general and applies to the ordinary federated model rather than to any particular company; the description of an intentions menu is of a product feature that the companies change whenever they like.

The donor-conception section contains no legal or medical advice, names no clinic and no route, and states only the shape of the framework and the fact that it has moved. Anybody with a question about her own situation is sent to a specialist and a lawyer.

The readings that are ours are marked as ours: that the agency’s real product is a stopping rule, that the intentions menu changed what was sayable before it could have changed behaviour, and that the market sells alternatives and satisfaction while ignoring the investment term. The commitment model, the maximiser finding and the review of online dating are other people’s work, reported with their limits and not stretched past them.