Modern Dating Statistics: The Numbers That Hold Up, and the Ones Nobody Publishes

Three kinds of dating statistic, and only two of them are evidence

Almost every number that circulates about modern dating comes from one of three places, and knowing which one a figure came from tells you most of what you need to know about it.

Counts. The Census Bureau's Current Population Survey records who is married, at what age, and who lives with a partner. The Bureau of Justice Statistics runs a national survey of crime victims. Nobody is asked how they feel about anything; people are counted. These are the most boring numbers in this genre and the hardest to argue with.

Independent surveys. Pew Research Center, the How Couples Meet and Stay Together study run out of Stanford, the General Social Survey, SSRS. They sample the population deliberately so that the results generalise, they publish their fieldwork dates and sample sizes, and they have no product to sell. They measure attitudes and experiences, which makes them softer than counts, but they can be checked.

Platform and commercial surveys. App companies, dating coaches and brands poll their own users or opt-in panels, publish the headline and rarely the method. Some of these are perfectly honest. The trouble is that from the outside there is no way to tell the honest ones from the rest, and they supply a large share of the percentages that end up on list pages of dating statistics, where a figure acquires authority simply by being repeated. It is the same way the three-day texting rule acquired its number.

This page used to mix all three without saying which was which, and worse. Until 10 September 2026 it carried figures attributed to Pew that Pew never published, and a Hinge report and a Match Group funnel that no public source contains. That history is on our corrections page.

What follows uses the first two kinds of source and labels the third where it appears. Every figure has its source linked at the bottom of the page, and where a number comes with a limit that changes what it means, the limit is in the same paragraph as the number rather than in a footnote.

How couples meet: the one question that has been measured properly

The best evidence on how couples meet comes from sociologists, not from an app. Michael Rosenfeld, Reuben Thomas and Sonia Hausen used two waves of the How Couples Meet and Stay Together survey, 2,495 respondents with heterosexual partners from 2009 and another 2,997 from 2017, to chart how couples met by the year they met.

The headline, published in PNAS in 2019: about 39% of heterosexual couples who met in 2017 met online, and meeting online overtook meeting through friends as the most common route around 2013. For same-sex couples who met in 2017 the figure was about 65%. In the 2009 data the heterosexual share had seemed to level off at about 22%; the smartphone years restarted the climb.

The paper's argument is in its title, Disintermediating your friends, and it rests on a detail most summaries skip. 89% of the couples in the 2017 survey who met online had no personal connection before they met. The internet did not just add a new venue. It took over a job that friends and family used to do, which was introducing people to each other.

Three limits, and each changes how the number should be read. It describes couples who formed, not people who are using apps: a route can be the most common way into a relationship and still be a frustrating experience for most of the people on it in any given month. It relies on people's own account of how they met. And it is a United States sample, so it says nothing directly about anywhere else.

There is also a category problem no survey fully solves. Plenty of people who say they met through a friend exchanged their first real messages on a phone, and plenty who met on an app would describe the first date as the moment they actually met. Met online and met offline were cleaner categories in 1995 than they are now, and every share in this section is softer than its rounding suggests.

How many people use dating apps, and how many are with someone they met on one

Two independent surveys, three and a half years apart, give the usage picture.

Pew Research Center surveyed 6,034 US adults from 5 to 17 July 2022. 30% said they had ever used a dating site or app, unchanged from the previous time Pew asked. Among partnered adults, meaning married, cohabiting or in a committed relationship, one in ten said they had met their partner on a dating site or app. That rose to 20% of partnered adults under 30, and to about a quarter of lesbian, gay and bisexual partnered adults.

SSRS surveyed 2,012 US adults in January 2026 through its Opinion Panel, with a margin of error of plus or minus 2.5 points. 37% had used a dating site or app at some point: 51% of 18- to 29-year-olds and 53% of 30- to 49-year-olds, against 20% of those 50 and older, and 63% of LGBTQIA+ adults against 34% of everyone else. But only 6% were using one at the time of the survey, and 10% of the youngest group.

Two of these numbers get confused constantly, so it is worth putting them side by side. SSRS found that half of the people who had ever used a dating app said they had at some time been in a committed relationship with someone they met through one. Pew found that one in ten partnered adults are with someone they met on one. Those are not in tension. The first is about everyone who has ever tried an app, across their whole lives, and counts any committed relationship, including ones that have ended. The second is about current couples only. Neither is a success rate, and neither tells you what share of the people using an app this month will leave it with a partner.

The gap between 30% and 37% is also not, on its own, proof that use rose between 2022 and 2026. They are two different firms using two different panels, and a gap that size between separate surveys is not a trend. What both agree on is the shape: common among adults under 50, most common among LGBTQ adults, and at any given moment used by a small minority of the population rather than by everyone who is single.

Being on your phone is not the same thing as being on an app, and what phones do to the people you are already sitting with is a separate and better-measured question.

What the companies publish, and what they do not

Match Group, which owns Tinder and Hinge, is a public company, which means some of its numbers are not marketing. They are filed with the US Securities and Exchange Commission.

Its annual report for 2025 gives the number of payers for each brand, in thousands: people who paid in a given month, averaged across the year, with the group total able to count the same person twice if they pay for two brands. Tinder: 9,026, down from 9,696 in 2024 and 10,375 in 2023. Hinge: 1,801, up from 1,532 and 1,242. Across the whole group, 14,165, down from 14,898 and 15,602. Total revenue for the year was about $3.49 billion.

That is real information, and it is the wrong information for the question most people bring to it. A payer is someone who pays. A falling count can mean people giving up, people finding someone and leaving, people moving to a different app, or people deciding the free version is enough, and the filing cannot tell those apart. A rising count, like Hinge's, can mean the product is working or that it is getting better at charging. Nothing in the report says how many users found a relationship.

That is the pattern across the industry. The companies report to regulators how many people pay; they publish trend reports about dating culture built on surveys of their own users; none of them publishes auditable data on what share of their users end up in a relationship because of the product. That absence is why figures like a match-to-date conversion funnel are so easy to invent and so hard to disprove, and it is exactly the gap this page fell into before its correction.

The one part of the proposition that has been tested independently is the matching, and it is the weakest part. When more than a hundred self-report measures collected before a speed-dating event were given to machine-learning models, they could predict how much a person tended to desire others and how much others tended to desire that person, and nothing at all of the part specific to a particular pair, which is the part any algorithm is selling. What that result leaves standing is in what questions can and cannot reveal about compatibility, and the rest of the case — what these apps do succeed at, and the measurement studies showing how unevenly the odds are spread — is in whether dating apps actually work.

Do relationships that start online go differently?

Two studies have looked at this with large samples, and the most-cited one comes with a disclosure worth reading first.

Cacioppo and colleagues (2013), in PNAS, analysed a nationally representative sample of 19,131 people who married between 2005 and 2012. The survey was run by Harris Interactive and commissioned by eHarmony, and one of the authors was a former director of eHarmony's research laboratory; the paper reports that two independent statisticians oversaw and verified the analyses. More than a third of the marriages had begun online.

Marital break-up, meaning separation or divorce, was 5.96% among those who met online and 7.67% among those who met offline. Among those still married, satisfaction averaged 5.64 against 5.48 on seven-point scales, a difference that shrank to 0.16 after controlling for demographic differences between the groups. The authors' own description of the size of those effects is that they are understandably quite small and do not hold for everyone.

Rosenfeld (2017), in Sociological Science, followed the 2,669 How Couples Meet and Stay Together respondents who answered at least one follow-up survey after 2009. Meeting online did not predict breakup. What it did predict, for heterosexual couples who met through dating websites, was a faster transition to marriage. Rosenfeld states plainly that he does not claim to measure any causal effect of the technology.

Put the two together and the defensible conclusion is modest. There is no evidence that relationships which start online are worse, and the evidence that they are better amounts to a small difference in one industry-commissioned survey. The reason to be cautious in both directions is selection: people who meet partners online differ from people who do not, Cacioppo's own data found demographic differences between the two groups, and no study can randomly assign someone to meet their partner on an app.

Once two people are together, how they met recedes quickly as an explanation for anything. What predicts how it goes from there is what happens between them, which is where the research on couples who stay together picks up.

Partnering later, and not quite in the straight line the headlines draw

The demographic direction is the part of the dating-is-broken story with the strongest evidence behind it, and it is less one-directional than it usually sounds.

Age at first marriage keeps rising. The Census Bureau's estimated median was 24.7 for men and 22.0 for women in 1980, 26.8 and 25.1 in 2000, and 30.8 and 28.4 in 2025. In 1956 it was 22.5 and 20.1. The 2025 row is flagged for a change in the population estimates underneath the survey, so single-year movements are not worth reading into, and women's median actually went down from 28.6 in 2024. The long trend is not in doubt.

More adults were unpartnered, and then slightly fewer. A Pew Research Center analysis of census data found that 38% of adults aged 25 to 54 were neither married nor living with a partner in 2019, up from 29% in 1990, and that all of the growth came from people who had never married. A later Pew analysis then found the share of all adults who were unpartnered fell from 44% in 2019 to 42% in 2023, with the decline showing up in every age group. Anyone quoting the first number as a still-accelerating trend is a few years out of date.

Divorce is not rising. The National Center for Family and Marriage Research at Bowling Green State University puts the refined divorce rate, divorces per 1,000 married women, at 14.4 in 2023, down slightly from 14.6 in 2022.

What none of this measures is quality. Later marriage could mean more careful choices or more difficulty finding anyone, and the census does not ask which. The same figures are read as a crisis by one commentator and as a sensible delay by the next, and the data cannot settle it for either of them.

If your own question is less about the national trend than about whether one relationship is heading somewhere, what actually signals that a new relationship has potential is the more useful page, and the 3-6-9 rule is why a calendar is the wrong instrument for answering it.

Less sex and less dating among young adults, measured two different ways

Ueda, Mercer, Ghaznavi and Herbenick (2020), in JAMA Network Open, used ten waves of the General Social Survey from 2000 to 2018, with 4,291 men and 5,213 women aged 18 to 44 in the analysis of sexual frequency. Among men aged 18 to 24, the share reporting no sexual activity in the past year rose from 18.9% in 2000-2002 to 30.9% in 2016-2018. Among men aged 25 to 34 it doubled, from 7.0% to 14.1%, and among women aged 25 to 34 it rose from 7.0% to 12.6%. Among women aged 18 to 24 the rise, from 15.1% to 19.1%, was not statistically significant.

That is a trend measured the same way across nearly two decades, which is what makes it worth more than a one-off poll. Its limits: survey response rates ran between about 60% and 71%, the data stop in 2018, and self-reports about sex depend on how comfortable people are answering the question.

The second measurement is newer and it is a snapshot. State of Our Unions 2026, published by the Institute for Family Studies and written by Alan Hawkins, Brian Willoughby, Jason Carroll and Brad Wilcox, draws on a 2025 national survey of 5,275 unmarried adults aged 22 to 35. About 30% said they were dating, casually or exclusively. 31% were dating once a month or more, 26% of women and 36% of men, while 51% said they were interested in starting a relationship. The report calls this a dating recession.

That word deserves a caution. A recession is a decline, and a single survey measures one moment: this one can show how much young adults date now, and cannot show that they date less than they used to. The numbers on this page that do show change over time are the General Social Survey ones above, and they point in the same direction, but they measure sexual activity, not dating, and they end eight years ago.

The gap the report does document is inside a single moment: half of the sample interested in a relationship and fewer than a third dating at all. A lot of that gap is people who are unsure how to start, and some of it is people in arrangements they cannot name, which is the subject of what has actually been measured about situationships.

Is dating harder? What people say, and the harassment numbers underneath

47% of Americans said in 2019 that dating had become harder over the previous ten years, against 19% who said easier and 33% who said about the same, in a Pew survey of 4,860 adults fielded from 16 to 28 October 2019. Women were much more likely than men to say harder, 55% against 39%. Asked in their own words why, 21% pointed to increased risk, ahead of technology at 12%, and women were twice as likely as men to give that answer.

That is a perception, and perceptions of whether something was easier ten years ago are not reliable. But the risk part of it has a measured counterpart.

In Pew's 2022 survey, 48% of people who had ever used a dating site or app had experienced at least one of four unwanted behaviours: being sent sexually explicit messages or images they had not asked for, being contacted again after saying they were not interested, being called an offensive name, or being threatened with physical harm. Among women under 50 who had used them, it was 66%, and 56% had been sent explicit content they did not want. Asked about safety, US adults split almost evenly: 48% said dating sites and apps were a very or somewhat safe way to meet people and 49% said they were not.

This is the most under-reported number in the whole subject. Arguments about whether apps work are almost always about yield: matches, dates, relationships. For a large share of the women using them, the defining experience is not low yield but unwanted contact, and that is a separate problem from whether the matching is any good. Fixing one would not fix the other.

If contact that started on an app has turned into monitoring, threats or pressure, it is no longer a dating problem. What controlling behaviour looks like and where the line falls names it plainly, and the helplines at the top of this page are free.

Loneliness, and what the Surgeon General's number rests on

In 2023 the US Surgeon General issued an advisory titled Our Epidemic of Loneliness and Isolation. Its most quoted line is that approximately half of US adults report experiencing loneliness, and it is worth knowing what sits under that sentence.

The advisory cites three sources for it. The first is a report by the insurer Cigna. The second is the peer-reviewed version of a Cigna-sponsored panel survey: Bruce and colleagues (2019), 20,096 US adults surveyed online in early 2018, who measured loneliness on the UCLA Loneliness Scale, which runs from 20 to 80. The paper reports an average score of 44.03. An average score is not a share of people. Getting from a scale to half of adults are lonely requires choosing a cut-off, and where the cut-off sits decides the headline. That does not make the concern wrong. It makes the precise fraction far less precise than it sounds.

The same paper found that being in a couple was associated with lower loneliness, and that the strongest associations were with social support and with meaningful daily interactions, while self-reported overuse of social media had a much smaller one. Those are associations measured at one moment, not causes.

A sturdier measure of the underlying change comes from how people actually spend their time. Kannan and Veazie (2023) used the American Time Use Survey, in which people record what they did across a day. Time spent with friends fell from 60 minutes a day in 2003 to 34 in 2019, and to 20 in 2020. Time spent alone rose from 285 minutes a day to 309, and then to 333. The 2020 figures are pandemic-year figures and should not be read as the new normal, but the decline from 2003 to 2019 predates the pandemic entirely, and the authors describe time with friends as having plummeted for the youngest Americans.

This connects to dating more directly than it looks. Friends used to be how couples were introduced; the time-use data show less time spent with friends in which that could happen, and the How Couples Meet data show apps taking over the introduction. The two findings fit together. Neither study was designed to prove that one caused the other, and nobody has.

Three claims this page used to make without a source, now checked

When this page was corrected on 10 September, three claims came out because nothing in it supported them: that relationship satisfaction has not declined, that intimate partner violence has fallen, and that relationships are more equitable than they were forty years ago. Here is what the primary sources say about each, and one of them turns out to need a qualification the old page never had.

Marital quality did not obviously decline, as far as it was measured. Amato, Johnson, Booth and Rogers compared two national surveys of married people, one from 1980 and one from 2000. Marital happiness and divorce proneness changed little over those twenty years, while marital interaction declined significantly. They found forces pulling in opposite directions: more economic resources, more equal decision-making and less traditional gender attitudes went with higher marital quality, while longer working hours and more demanding jobs went with lower. The data end in 2000, so this is evidence against a long decline up to then, not a statement about 2026.

Intimate partner violence fell steeply, and has risen since 2020. The Bureau of Justice Statistics reported that from 1994 to 2010 the rate of intimate partner violence fell by 64%, from 9.8 to 3.6 victimisations per 1,000 people aged 12 or older. That is the fact behind the old claim, and it is a large one. But the Bureau's most recent annual report shows the rate rising from 1.7 per 1,000 in 2020 to 2.7 in 2024, with no statistically significant change between 2023 and 2024. The honest version is that the long-run fall is real and the last few years have gone the other way. The old sentence, which said researchers consistently find it has fallen, would have been out of date even with a source attached.

Marriages are more equal in earnings, and less equal at home than that suggests. In a Pew analysis of opposite-sex marriages, both spouses earn about the same, each between 40% and 60% of the couple's combined income, in 29% of marriages today. The share where the husband is the sole provider fell from 49% in 1972 to 23%, and the share where the wife is the sole or primary earner rose from 5% to 16%. Yet in the marriages where earnings are equal, wives spend about two hours more a week on caregiving than husbands and about two and a half hours more on housework. More equitable on one dimension, then, and measurably not on another.

None of the three is a verdict on whether modern relationships are better. They are separate measurements, and the reason they were worth checking is that each had been flattened into a reassuring sentence the sources only partly support.

What the numbers are good for, and what they cannot tell you

After all of that, a short list of what is actually established. Meeting online became the most common way couples in the United States meet, around 2013. Somewhere between three and four in ten adults have used a dating app, and a small fraction are using one at any given moment. Couples who met online do not break up more often. People marry for the first time years later than they did in 1980. Young men reported far more sexual inactivity in 2018 than in 2000. Time with friends was falling well before the pandemic. And a large share of the women who use dating apps have been harassed on them.

What is not established is almost everything people actually want to know. There is no public success rate for any app. There is no measure of whether relationships formed now are better or worse than relationships formed in 1990. There is no study connecting any of these national trends to what will happen between you and one particular person.

That last point is not a disclaimer. It is the most important thing on the page, because averages are where this genre goes wrong once people use it to make decisions. Knowing that a third of adults have used an app tells you nothing about whether you should, and knowing that the median man now first marries at 30.8 tells you nothing about when you will.

The questions that do have answers at the level of one relationship are about what happens inside it: how conflict goes, whether you feel appreciated, what you do when you are anxious about where you stand. The attachment style test and the relationship satisfaction test are built for that level, and both say plainly what a score cannot tell you. Why intelligent people keep choosing badly is the same question asked about repetition. More on this subject under Dating.