The state of connection

How do most couples meet? What the Stanford data actually says

By Matthew Firth, founder of NexSpark · Aug 7, 2026 · 8 min read

Search this question and you will be told 39%, 50%, 60% and 10%, all with confidence, often on the same page of results. Most of those numbers trace back to one Stanford survey, and most of the articles quoting it have misread it in the same three ways.

The survey is How Couples Meet and Stay Together, run by the sociologist Michael Rosenfeld. The headline finding was published in the Proceedings of the National Academy of Sciences in 2019 under a title that gives away the argument: Disintermediating your friends. It is the most-cited source on this question, and it is worth reading rather than reading about.

The short answer

Among heterosexual couples in the United States who met in 2017, 39% met online. That was more than any other single way of meeting. Bars and restaurants show 27% and friends 20%, but as the next section explains, most of that 27% is couples who met online and then went to a bar, so friends is the real second.

Here is the full table from the paper, comparing couples who met in 1995 against couples who met in 2017.

Online passed friends around 2013. Before that, from the end of the Second World War onward, being introduced by a friend was the most common way American couples met, and it held that position for about sixty years.

Same-sex couples got there first. Roughly 65% of same-sex couples who met in 2017 met online, and that share has been broadly stable since 2009. Online dating solved a thinner market earlier.

Three things almost everyone gets wrong

1. The bar figure is an artefact

Bars and restaurants are the only category besides online that rose, from 19% to 27%, and that number gets quoted constantly as evidence that going out still works.

The paper is explicit that it does not mean that. The rise is due entirely to couples who met online and then had their first in-person meeting at a bar or restaurant. Strip those couples out and bars and restaurants have been declining as a place to meet since 1995.

So the bar is not where couples meet. It is where the app sends them. The 27% is a measurement of first dates, not of first meetings.

2. The percentages do not add up, and are not supposed to

Add that column and you get well over 100%. That is not an error. The categories are not mutually exclusive, because people were asked to tell the story of how they met in their own words and every applicable category was coded. A couple introduced by a friend who then messaged on Facebook counts in both rows.

This matters because it makes every pie chart of this data wrong, and there are a great many pie charts of this data. A pie chart asserts that the slices are exclusive and sum to the whole. Neither is true here.

3. The college decline is not a real finding

Meeting in college goes from 9% to 4%, and gets written up alongside the rest as another casualty of the apps. It is the one row in the table that is not statistically significant. The authors say so directly, and attribute the apparent drop to sparse data in the final years. Every other row in the table cleared significance, though bars and church only just. This one did not clear it at all.

Why one source says 39% and another says 10%

Pew Research, in a survey fielded in July 2022, found that 10% of partnered adults met their partner on a dating site or app. Rising to 20% among partnered adults under 30, and about a quarter among partnered LGB adults.

That looks like it contradicts 39%. It does not. The two numbers answer different questions.

• 39% is a flow. It describes couples who formed in one recent year.
• 10% is a stock. It describes every partnered adult alive today, including people who met in 1975, when online dating did not exist.

Most long relationships began before the apps, so they hold the average down permanently. If you want to know how people are meeting now, you want the flow. If you want to know how the couples around you met, most of whom got together years ago, you want the stock. Quoting one to debunk the other is a category error, and it happens a lot.

The number that is nine years old

Worth saying plainly, because almost nobody does: the 39% is from a survey fielded in July 2017.

There are later waves, in 2020 and 2022, and they are genuinely useful, but they are follow-ups. The sampling frame is the same people who answered in 2017 and who were still on the panel. They track what happened to those couples. They are not a fresh, nationally representative sample of how couples met.

Which means that as of today there is no newer national answer to this question than 2017. Anyone citing 39% as a current figure, including the sites currently ranking for this search, is citing a nine-year-old survey without saying so. The direction of travel since then is almost certainly the same. The precise number is unknown.

The finding that nobody quotes

The share who met online is the headline. It is not the most interesting row in the study.

Rosenfeld also looked at how people met online, and found something sharper. Among couples who met online:

• The proportion who were complete strangers beforehand rose from 81.0% to 89.5%.
• The proportion introduced online by a third person collapsed from 11.2% to 3.7%.

Set that beside meeting through friends falling from 33% to 20% and the pattern is not really about technology. It is about the removal of the middleman. Friends and family used to do a specific job before you ever met someone: they knew both of you, they had a view on whether it made sense, and their own standing was on the line if it went badly. That job has quietly disappeared. Nine out of ten people who meet online meet a stranger, with nobody in between who knows either of them.

The paper calls this disintermediation, and compares it to what the internet did to travel agents. The comparison is apt and slightly bleak. Booking your own flights is fine. Vetting your own strangers, at volume, with no help, is the thing people are describing when they say dating is exhausting.

What we took from it

I will be straightforward about the interest here, since this is a matchmaking company's website. The disintermediation finding is the reason NexSpark exists in the shape it does.

The answer we landed on is not nostalgia. Going back to waiting for a friend to introduce you is not a plan, and the data shows why: that channel has been shrinking for thirty years and shows no sign of turning. Friends are not a scalable introduction system and never were, which is precisely why something replaced them.

What we do instead is put the intermediary back on purpose. Every member is identity-verified, so the person across the table is who they said they were. A real person reviews every introduction before it is made, so somebody has actually considered whether the two of you make sense. That is the function friends used to perform, rebuilt deliberately rather than left to chance.

The other half of that collapse is not about dating at all. Friend networks did not shrink only as a route to romance; they shrank as social infrastructure. That is a separate problem, and it is the one NexSpark Friends was built for.

Common questions

How do most couples meet today?

Among heterosexual couples who met in 2017, 39% met online, more than any other single way. Meeting through friends was the real second at 20%, once the bar and restaurant figure is corrected for couples who met online first. Bars and restaurants show a higher raw figure, 27%, but most of that is couples who met online and then arranged to meet at a bar. Those figures come from Stanford's How Couples Meet and Stay Together survey, and 2017 is still the most recent nationally representative sample of how couples met.

Why do some sources say 39% of couples met online and others say 10%?

They are answering different questions. The 39% describes couples who formed in a single recent year. The 10%, from Pew, describes all partnered adults alive today, including people who met in 1975 before online dating existed. Both are correct. Neither is a good check on the other.

Do 27% of couples really meet in bars and restaurants?

No. Bars and restaurants appear to rise from 19% to 27%, but the researchers state that the rise is due entirely to couples who met online and then had their first in-person meeting at a bar. Excluding those couples, bars and restaurants have been declining as a place to meet since 1995.

What is the most common way couples met before online dating?

Friends. From the end of the Second World War until around 2013, being introduced by a friend was the single most common way heterosexual American couples met. It peaked at roughly a third of couples and has fallen to about a fifth.

How do same-sex couples meet?

About 65% of same-sex couples who met in 2017 met online, against 39% of heterosexual couples. Same-sex couples adopted online meeting far earlier, and that share has been roughly stable since 2009.

Sources

Michael J. Rosenfeld, Reuben J. Thomas and Sonia Hausen, Disintermediating your friends: How online dating in the United States displaces other ways of meeting, Proceedings of the National Academy of Sciences 116(36), 2019. Table 1 and Table 2. Survey data from How Couples Meet and Stay Together, 2017 wave, fielded July 2017, 3,510 respondents, of whom 2,997 described meeting a different-sex partner. The paper names GfK as the survey firm; the panel was later acquired by Ipsos, which is how Stanford's own documentation now refers to it. Wave structure confirmed against the HCMST 2017 to 2022 user's guide. Pew figures from Key findings about online dating in the U.S., Pew Research Center, published February 2023 from a survey fielded 5 to 17 July 2022.