The Arranger Who Profits From the Search
America, 2026 — an essay on choice, capital, and the algorithmic return of arranged marriage
“The freedom to marry has long been recognized as one of the vital personal rights essential to the orderly pursuit of happiness by free men.”
— Chief Justice Earl Warren, Loving v. Virginia, June 12, 1967
“Significant value creation.”
— Elliott Investment Management, on its stake in Match Group, 2024
I. The Promise
The West built its postwar self-image on a claim about the most intimate decision a person makes. Marriage, it said, had stopped being a family’s transaction, a community’s settlement, or a state’s instrument. It had become a choice — made by two people, for reasons that were theirs.
The genealogy is real. Companionate marriage displaced arranged marriage as the dominant ideal across the twentieth century. Loving v. Virginia struck down the last laws criminalizing interracial marriage in 1967, folding romantic choice into the civil-rights settlement. California’s Family Law Act of 1970 opened the era of no-fault divorce and completed the picture: marriage as a contract free individuals could enter and leave on their own terms.
The whole architecture assumed one thing above all else — that the two people involved would be the ones doing the choosing.
By 2017, according to survey data analyzed by the Stanford sociologist Michael Rosenfeld and colleagues, meeting online had become the single most common way American heterosexual couples met, at roughly 39 percent — ahead of friends, family, work, and school. That is not every marriage, and this essay does not claim it is. It is the plurality channel: the largest single mechanism by which Americans now find the people they marry. And that channel has an arranger.
II. The Inverted Arranger
Every older system of mediated matchmaking shared one structural feature. The family negotiating a marriage, the community making an introduction, the professional matchmaker earning a referral — each had a stake in the match lasting. Reputation, kinship, and repeat business all pointed the same way.
The platform’s stake points the other way. A user who marries and leaves is a canceled subscription. That is not an imputation of motive; it is the revenue model, and the public record shows what the model does under pressure. The Federal Trade Commission’s 2019 complaint against Match Group alleged that the company sent “you have a match” notifications to non-paying users from accounts it had already flagged internally as likely fraudulent, and that nearly 500,000 subscriptions were purchased within a day of such notices between 2016 and 2018. The case settled in 2025 for $14 million, without any admission of liability.
The science behind the product is no firmer. A 2012 review in Psychological Science in the Public Interest, led by the Northwestern psychologist Eli Finkel, found no compelling evidence that any dating site’s matching algorithm predicts long-term relationship success — the systems measure what is easy to measure, not what relationship science says matters. A 2025 paper in the Journal of Applied Philosophy went a step further, asking whether users have a right to know if an app is designed to prolong their search rather than conclude it.
This is the pattern the Review has elsewhere traced through addiction treatment and the casino floor: an industry whose business requires a permanent population of the managed-but-not-fixed. A resolved customer is a lost customer. The dating app is that architecture applied to the heart.
III. The Masands, Again
The Sikh tradition has a precise name for what happened. Guru Nanak’s revelation was direct and unmediated — no priesthood, no gatekeeper between the seeker and the Divine. As the community grew, the masands arose as administrative infrastructure to collect offerings and carry teaching across distance. They became, in time, the very gatekeepers the revolution had abolished. Guru Gobind Singh’s answer was not a better class of masands. It was to end human mediation altogether and vest authority in a fixed text, the Guru Granth Sahib — something no future administrator could quietly rewrite. This is the Masands Cycle.
The postwar choice revolution is the Nanak moment: a genuine liberation of intimate life from family and community control. The platform is the masand that arose once matchmaking outgrew word of mouth — useful infrastructure first, a new gatekeeper after.
What makes this case darker than the historical one is the missing resolution. There is no text behind the algorithm. A match percentage can be redesigned tomorrow, in private, by whoever controls the company that quarter. The Review’s reading of the Wachowskis in The Matrix Bridge draws the relevant distinction: the Architect speaks in closed equations and hands Neo percentages; the Oracle hands him a cookie. One is calculation. One is prasad. “89% compatible” is Architect logic exactly — a number offered in place of an encounter, with no one standing behind it who can be asked to account for it.
The person swiping is not a dupe. In the Gurmat vocabulary, they are a manmukh — a consciousness handed real resources and real agency, spending both inside a system it does not perceive, while haumai, the ego-self, supplies the sincere feeling of having chosen. The feeling is authentic. The choice was arranged.
IV. The Artificial State
Yuval Noah Harari’s Homo Deus names the deeper stake. Liberal humanism rests on one premise: that the individual’s own feeling is the final court of appeal. That premise underwrites the ballot — your vote reflects your judgment — and it underwrites postwar marriage in the same breath: your heart reflects your judgment. Systems that predict and steer feeling better than people can introspect it displace that premise in both places at once.
That licenses a stronger word than metaphor. Hannah Arendt’s account of totalitarianism in The Origins of Totalitarianism does not require a dictator or a camp. It requires the atomization of the intermediate structures — family, congregation, kin — that stand between the isolated individual and total power, and the collapse of the boundary between public and private life, because total domination is defined by how far into intimate life it reaches.
The current literature on algorithmic authoritarianism concentrates almost entirely on the public sphere: surveillance, censorship, predictive policing, election manipulation. By Arendt’s own definition, that account stops short. A critique of algorithmic power that never reaches mate selection and family formation has left out the sphere her definition treats as decisive.
Sheldon Wolin’s inverted totalitarianism supplies the name for what results — domination without a visible ruler or a single ideology, produced through corporate architecture and a consumerized public. Shoshana Zuboff’s instrumentarian power supplies the mechanism: control that needs no believers, only predictable and modifiable behavior, produced here by the variable reward of the swipe.
V. Full Circle
None of the functions this system performs are new. What changed is who administers them.
Paternity. Friedrich Engels argued in The Origin of the Family, Private Property and the State that monogamous marriage existed to secure inheritance through certain paternity, enforced for most of history through surveillance of women’s bodies. Consumer DNA testing is the same anxiety, privatized.
Child-rearing. Sarah Blaffer Hrdy’s research on cooperative breeding describes the human default as children raised across dense kin networks. The isolated nuclear household that replaced it is overseen not by grandmothers and aunts but by child protective agencies — a surveillance function the village once performed informally, now performed by the state through coercive channels.
Mate selection. Here the American precedent is uncomfortably direct. From the eugenic marriage statutes of the 1910s through Connecticut’s premarital blood-test law of 1935, states made medical certification a condition of a marriage license, framed as protecting future offspring; by mid-century nearly every state required it, and the last such requirement was repealed only in 2019. That regime was public and statutory — a legislature passed it, and legislatures ended it. What operates now is private and proprietary, with no repeal mechanism, because no electorate ever voted it in. Reporting in Time on Match Group’s patent filings described relevance systems that sort by traits including hair color, eye color, and ethnicity.
The connecting thread is what Max Weber called rationalization: communal and traditional authority displaced by legal and then technical authority. The functions never left. The form of authority administering them changed three times, in one direction.
VI. The Money Trail
The arranger has owners, and their incentives are on file with the Securities and Exchange Commission.
Match Group — Tinder, Hinge, OkCupid, Match.com, and roughly two dozen other brands — was assembled inside Barry Diller’s IAC, which bought Match.com for about $50 million in 1999 and spun the portfolio off as an independent public company in 2020. In early 2024, the activist fund Elliott Investment Management disclosed a stake of roughly $1 billion and described its aim as “significant value creation.” Within a year, chief executive Bernard Kim was out after a sustained decline in paying Tinder users, and Spencer Rascoff, a Zillow co-founder who had joined Match’s board during the activist campaign, replaced him. The metric at the center of the transition was payers — not marriages, not relationships formed. Payers.
Bumble has been controlled by Blackstone since a 2019 deal valued at about $3 billion bought out founder Andrey Andreev’s interest. Its class structure gives Blackstone voting power well above its economic stake. By 2025 the company was reporting double-digit declines in revenue and paying users, and founder Whitney Wolfe Herd, back as chief executive, was rebuilding the product as “AI-first.”
Both companies answered documented decline the same way: more of the mechanism. Match Group has committed tens of millions of dollars to a centralized AI platform, including conversation-practice features built on OpenAI’s models. That places the arranger on a larger rail — the AI infrastructure buildout in which Nvidia invests in OpenAI, OpenAI buys compute from Microsoft, Oracle, and CoreWeave, and those firms buy chips from Nvidia. The Bank for International Settlements and the Financial Stability Board have both flagged the concentration and circularity of that buildout as a risk to financial stability.
Above all of it sit BlackRock, Vanguard, and State Street, which together manage more than $20 trillion and, in one widely cited study, rank as the largest combined shareholder in nearly nine of ten S&P 500 companies. They hold the dating apps. They hold the chipmaker, the clouds, and the model labs beneath them. As index funds, they rarely sell and usually vote with management. Economists call this common ownership: an owner holding every competitor has little reason to force any one of them toward a costly correction.
Everything above is drawn from SEC filings, court dockets, and published reporting, and nothing above is an accusation against any named person. The FTC matter settled without admission of liability. No executive or investor named here is alleged to have acted outside the incentives the structure hands them — which is the argument. A mechanism that requires villains can be fixed by replacing them. This one cannot.
VII. The Successor
The inverted arranger has one structural weakness, and it is showing in the filings. A model that monetizes the search still requires the searcher. The declining payers, the burnout majorities, the billion-dollar activist stake demanding “significant value creation” — this is an industry exhausting the population it manages. A user who despairs of the search churns as surely as one who marries.
The industry’s answer is visible in the same filings, and it is not a better search. It is the end of the counterpart.
In October 2025, OpenAI announced that ChatGPT would permit erotica for verified adults beginning that December — the largest AI laboratory in the world formally entering the intimacy market, on the stated principle of treating adult users like adults. The dedicated companion platforms — Replika, Nomi, Candy AI, and a dozen rivals — were already there, selling not access to possible partners but the partner itself: persistent memory, synthetic affection, attachment engineered to retain. The AI-companion market was estimated near $37 billion in 2025, projected to compound around thirty percent annually — already several times the size of the entire dating industry, growing several times faster.
The inversion completes itself here. The dating app monetizes hope: revenue while the search continues. The companion monetizes satisfaction: revenue while the attachment holds. The first loses its customer to a wedding; the second can lose its customer only to disillusion — the one contingency the product is optimized, daily, against. The arranged system resolves its founding contradiction the only way available to it: by abolishing the second party.
The Masands frame names what this is. Guru Gobind Singh ended mediation by vesting authority in a fixed text no administrator could rewrite. The companion is the precise inverse — a mediation with nothing fixed behind it, rewritten continuously by the administrator, in private, toward retention. The masand no longer carries the offering across distance to the Divine. He generates the Divine on the premises, and meters the visit.
And the agency is leaving the user as well. Bumble — Blackstone-controlled, rebuilding “AI-first” — is phasing out the swipe in favor of an AI assistant that matches on the user’s behalf. Startups already offer agents that build the profile, write the messages, and negotiate with the counterpart’s agent; the humans are introduced after the software has agreed. The arranged marriage this essay has traced as metaphor completes itself as procedure: two arrangers, negotiating a match, between principals who meet when the terms are set. The families are now software.
What becomes scarce, at the end of this road, is the one thing the system cannot synthesize: a verified human. A third of surveyed singles already say they would accept active monitoring of their conversations; nearly half support periodic re-verification. The premium tier of the next industry writes itself — the certified person, the organic produce of the relationship market — and the fraud economy pressing from the other side (romance-investment scams alone took $8.6 billion in a single year, per the FBI) guarantees the demand. Every agent-driven scam that burns a dater is an acquisition channel for the companion. The two industries do not merely converge. They feed each other.
VIII. The Law That Chose Not to Look
A country that broke up Standard Oil in 1911 and AT&T in 1982 has not touched any of this. The Sherman and Clayton Acts are still in force. What changed is the lens courts apply to them — and the change predates the technology by three decades.
Robert Bork’s The Antitrust Paradox (1978) argued that antitrust should ask one question: do consumer prices rise? Not market structure, not concentration, not power. Federal courts adopted that consumer-welfare standard through the late 1970s and 1980s, and it governs still. Tinder is free to join. Under a price test, there is no harm to show, however many nominal competitors sit inside one holding company.
The second barrier is market definition. In November 2025, Judge James Boasberg ruled against the FTC’s monopolization case against Meta after defining the relevant market broadly enough to include TikTok and YouTube. Put the same question to Match Group: is the market “dating apps,” where its share is dominant, or “every way people meet a partner,” where it disappears?
The third barrier has never been tested. Common ownership across an entire market sits outside developed antitrust doctrine. The theory exists in law reviews; no enforcement case has been brought. And the current FTC leadership has foregrounded platform “censorship” as its signature Big Tech concern. Dating markets are not on the agenda.
This is not the law failing to keep pace. Bork wrote before Tinder’s founders were in high school. The vulnerability was installed in advance.
IX. What Remains Possible
The record also shows two openings.
The first is doctrinal. The Clayton Act’s own text reaches mergers whose effect “may be substantially to lessen competition” — a preventive standard aimed at concentration before harm is proven, which the consumer-welfare era read down. Restoring that incipiency reading, or adopting something like the European Union’s abuse-of-dominance doctrine under Article 102, would let enforcers treat degraded quality, exploitative data extraction, and foreclosed alternatives as harms in their own right, independent of price. Neither requires invention. Both are existing law, somewhere.
The second is already working. In August 2026, Meta settled with a bipartisan coalition of state attorneys general for a sum reported between roughly $12 billion and $18 billion over ten years, depending on how its terms are counted, resolving claims that it designed Instagram to addict young users while documenting the harm internally. That case was never antitrust. It was consumer protection — a theory that asks only whether a design was harmful and whether the company knew. Whether it extends from adolescent mental health to adult relationship formation is an open legal question. It is no longer an unthinkable one.
The Masands Cycle suggests what a durable answer looks like. Guru Gobind Singh did not appoint better masands; he installed something they could not rewrite. The nearest institutional equivalent is not another lawsuit that must be won again each decade, but a bright-line rule — an ownership cap, a structural review, an interoperability mandate — that does not depend on winning the argument every time.
Until then, the arrangement continues. Somewhere tonight a person opens an app certain they are choosing. The feed refreshes. It is designed never to come up empty.