He Didn't Write That Opener. Neither Did She. Welcome to AI Dating in San Francisco.
Let's paint the scene.
It's a Wednesday evening. He's on Hinge from his Mission apartment, composing what reads like a relaxed, self-aware opener. References her hiking photo. A line about a taqueria on Valencia. Sounds like someone who has figured out how to be a person.
She reads it from her place in the Castro and thinks: finally, someone who doesn't open with a system design question.
ChatGPT wrote it. He is a senior engineer at a company fourteen blocks from OpenAI's office. He spent his workday evaluating the outputs of large language models. He spent his evening using a large language model to write a first impression. He did not find this ironic until right now, reading this sentence.
She, for her part, had Claude help her bio sound "warm but not thirsty." Her opening question was workshopped through Hinge's AI coach. Her profile was optimised with the same iterative, hypothesis-driven rigour she applies to product decisions at work. It performs well on the metrics she's tracking. She has not been on a second date in four months.
Two AI engineers. One 7x7 mile city. The most sophisticated AI dating infrastructure on earth. Zero working relationships.
Welcome to AI dating in San Francisco — the city that built the thing, can't make it work, and is genuinely surprised by this.
The Numbers First
San Francisco is, by almost every material measure, one of the most challenging dating environments in America.
The SF metro area's population in July 2025 was still 2.6% below its April 2020 level — making it one of only two major metros in the country that has not recovered from pandemic population loss, alongside Los Angeles.1 The AI boom that has remade the city economically has not, as of mid-2026, rebuilt its population to pre-pandemic levels. There are fewer people to date in San Francisco than there were five years ago.
The ones who stayed are paying for it. The median rent in San Francisco across all unit types stood at $4,050 per month in Q1 2026.2 The median house price has passed $2 million.3 Average date costs run $85 to $120 per outing.4
One in four San Francisco jobs are in tech.4 73% of SF singles report work as their biggest dating obstacle.4 78% use dating apps — a rate that local matchmakers describe, without qualification, as leading to "oversaturation and burnout."4 And on those apps, per Hinge's own data, men outnumber women 2 to 1.5
So: smaller population than five years ago. Eye-watering cost of living. Three in four singles say their job is the problem. Four in five are burned out on the apps. And a gender imbalance on the platforms that makes the whole situation structurally lopsided.
Into this, the engineers who built the AI are using the AI to date. Results pending.
The City That Built The Thing
Here is the specific San Francisco angle that no other city in any version of this story can claim: the models writing everyone's dating openers were, in significant part, designed, trained, and deployed within walking distance of where the people using them live.
OpenAI is here. Anthropic is here. Google DeepMind's applied research is here. The engineers who spent their days building the language models that are now writing romantic messages are, in many cases, the same people opening Hinge at 9pm and deciding whether to write the opener themselves or let the model do it.
This is not a hypothetical irony. In June 2025, the SF Standard reported on twin Princeton CS graduates — Jenny and Chrissy — who work in AI and created a viral AI-generated parody video asking the specific question the entire city has been circling: "If the gender ratio in San Francisco is supposedly so great, why are so many great women still single?"6
The video was AI-generated. About AI dating. Made by AI engineers. To describe their own AI dating problems. In the city that made the AI.
This is either the sharpest possible encapsulation of the problem or the most San Francisco thing that has ever happened. Possibly both.
The Optimization Mindset Cannot Date
San Francisco has a specific cultural pathology that predates AI and has been amplified by it, and local matchmakers have been naming it clearly for years.
A piece published in February 2026 put it directly: "People aren't products. Connection isn't a metric. And the relentless optimization mindset that makes you great at your job often makes you terrible at dating."7
This is the core of San Francisco's dating problem, stated as plainly as it can be. The city has produced a professional class of people who are exceptionally skilled at iterating on systems, improving metrics, and finding the optimal solution to a defined problem. It has produced the worst possible preparation for the actual experience of falling for another person.
Dating is not a funnel. A profile is not a product. A first date is not a sprint review. And yet: 45% of SF singles report meeting people through work only — which means nearly half the city's singles have effectively reduced their romantic search to their professional network, the one domain where they know how to optimize.4
The "Always Something Better" problem is documented and named. In a city built on disruption, there is always a better app, a better strategy, a better match one swipe away. As one SF matchmaker wrote: "You go on a good date, but you're wondering if you're settling. You meet someone great, but you keep your options open just in case. Commitment feels like closing doors, and in San Francisco, you've been trained never to close doors."7
AI does not solve this. AI is the most efficient machine ever built for keeping all doors open simultaneously. It lowers the cost of maintaining every option to near zero. An AI-assisted Hinge conversation costs almost nothing to sustain. Which means there is even less pressure to make a decision, close a door, or commit to something real.
The Bootcamp
In August 2025, thirty-four of the Bay Area's most eligible singles gathered in a Hayes Valley apartment for something called Hot Feelings — a hybrid communications masterclass and dating mixer created by Joyce Zhang, a product manager turned dating coach for high performers.
The attendees: startup founders. AI engineers. C-suite executives. The SF Standard described the vibe as "predictably awkward." No appetizers, no cocktails. Just water with lemon, shoes off at the door, and a commitment to be fully present for three hours. No phones.8
At the end of the evening, a private equity professional summarised what the room had collectively learned: "I learned that I don't know how to date."8
Others agreed.
This is a room of people who, by conventional measures, are among the most capable in the world. They have built companies, raised capital, shipped AI systems, and navigated some of the most demanding professional environments on earth. They cannot date. Not because they lack intelligence or resources or opportunity. Because dating requires a specific kind of presence, vulnerability, and tolerance for uncertainty that the professional culture of San Francisco has spent years training out of them.
AI providing a better opener does not address any of this. The opener was never the problem.
The Double Standard, Silicon Valley Edition
Nationally, 26% of singles use AI to enhance their dating lives — a 333% jump in a year, per Match and the Kinsey Institute's 2025 Singles in America report.9 Among Gen Z and millennials, the Hilly dating app found 82% and 87% respectively using AI for dating assistance — while 62% and 70% would lose interest discovering their match had done the same.10
In San Francisco, where a meaningful fraction of the dating pool has direct professional exposure to how these models work, the double standard takes on a particular quality. The engineer who knows exactly how the model generates text, who understands precisely what a well-prompted opener does and doesn't represent about the person who wrote it — is still using the model. And is still faintly annoyed to discover the other person did too.
A 2025 study found that people are only 57% accurate at identifying AI-written text — barely above chance.10 GPT-4.5 was mistaken for a human 73% of the time in direct conversation tests.10 In a city where the people taking these tests built the models being tested, the inability to detect AI-generated text is somehow even more pointed.
The Neighbourhood Problem Nobody Talks About
San Francisco is 7 by 7 miles. It is, on paper, one of the most compact major cities in America.
And yet: Marina people date Marina people. Mission people date Mission people. The East Bay might as well be a different state.7
The micro-geography of San Francisco is real and documented. The fog line that divides the Sunset from the rest of the city is a social divide as much as a meteorological one. The BART ride to Oakland is, for many SF singles, psychologically equivalent to a long-distance relationship. Noe Valley is a different social world from SoMa. The hills are not neutral.
In a 7x7 city, people have managed to recreate the suburban pod problem — self-contained social ecosystems that rarely interact, geographic loyalty that overrides the size of the dating pool, neighbourhood identity that functions as a compatibility filter before a single message is sent.
AI generates openers that cross these lines effortlessly. It does not generate the willingness to take the BART, or to try the bar in the neighbourhood you've never been to, or to spend a Sunday morning somewhere outside the social radius you've been rotating in since you moved here.
What The City That Built This Has Learned
Here is what San Francisco, more than any other city, should understand about AI dating — because it is the city that built the product and has been running the experiment longest.
Optimising the funnel does not improve the outcome if the problem is in the person, not the pipeline. Reducing friction does not help if the friction was the signal. Better openers do not produce better relationships if the humans behind them have been trained to optimise away the vulnerability that connection actually requires.
Jenny and Chrissy, the Princeton twins who made the viral video, said something that cut through all of it: "If you know to hold open the door, be a good listener — all these things — then you'll do very well."6
This is not a new insight. It is not an AI insight. It is the oldest available information about human relationships, expressed by two AI engineers in the city that invented the AI, in 2025.
The city that built the models knows what the models can't do. The question is whether it will act on that knowledge or keep optimising the thing that isn't the problem.
At Luvo, we've always known that connection isn't a metric. The city that built the AI figured it out too. We do the matching. You bring the humanity.
Sources:
Footnotes
SF Chronicle / Yahoo News, "SF's Exodus Still Hasn't Reversed," citing U.S. Census Bureau estimates through July 2025, March 2026 ↩
Relocity, "Bay Area Market Rental Trends Report," Q1 2026 ↩
Press Democrat, citing California Department of Finance data, April 2026 ↩
Met By Nick, "Dating in San Francisco 2025: The Complete Guide," citing U.S. Census ACS 2023, Bay Area Council Economic Institute Tech Employment Report 2024, SF Gate Dating Survey 2024 ↩ ↩2 ↩3 ↩4 ↩5
Reason Future Tech, "Is Dating in Silicon Valley Difficult?", citing Hinge platform data ↩
SF Standard, "An AI Video About SF Dating Was All Too Real: Meet the Twin Sisters Who Created It," June 2025 ↩ ↩2
Nick Rosen / Medium, "Finding Love in San Francisco: Why High-Achieving Singles Are Turning to Matchmakers," February 2026 ↩ ↩2 ↩3
SF Standard, "Killing It at Work But Losing at Love? Try This Bootcamp," August 2025 ↩ ↩2
Match & The Kinsey Institute, Singles in America 2025 ↩
Hily, T.R.U.T.H. Report 2025, via Fast Company; Scientific American, "The Rise of AI Chatfishing in Online Dating," October 2025; Norton Research, 2025 ↩ ↩2 ↩3