My AI Matchmaker Let Me Down

In the Instagram video, a knockoff JFK Jr. towers over a beautiful woman with a raspy voice. “Are you single?” the woman asks. He smirks into the camera. “I am.”

I had no choice: I clicked on the account, only to discover an endless stream of gorgeous single men looking for love. They were movie stars compared with the men I’d seen on dating apps. Where the hell was this woman finding these guys? And could I come?

The man-in-the-street interviews had been filmed by Amata, one of a handful of new AI matchmaking companies that present themselves as the future of online dating. Instead of scrolling through thousands of options, users are presented with potential matches one at a time. Instead of paying to use the app, you can choose to pay only to set up a date. (Amata charges $20 a date.) There’s less chatting: On Amata, the communication window opens only two hours before a scheduled meeting, saving the getting-to-know-yous for the actual date. One woman who uses Amata told me she liked the mystery of this; the dates feel more like old-fashioned setups. And the apps attempt to limit ghosting—they don’t refund the fee if you back out of a date.

But public trust in artificial intelligence in general is declining, even as the technology becomes woven into daily life. Searching for love is one of the most quintessentially human experiences. Will people really be willing to hand that search over to AI, along with information about their most intimate preferences and desires?

Maybe—if they’re desperate enough.

Last summer, a Forbes Health survey found that 78 percent of all dating-app users say they’re burned-out. Swiping through profiles isn’t romantic; at best, it’s a chore, and at worst, it’s a compulsion. Dating-app companies are especially worried about Gen Z. Young people who grew up online seem uninterested in finding love there. But the industry, which made $6 billion last year off the hopes of would-be lovers, isn’t about to give up on this market. Companies are betting that they can sell AI as an online-dating time-saver, and that people who are already depending on an algorithm to find love won’t be too spooked by this new technology. According to Ludovic Huraux, a co-CEO of Amata, “The future is AI matchmaking.”

[Annie Joy Williams: Canceled by Hinge]

Sitch, one of the first successful AI matchmaking apps, launched in 2024. One of its founders, Nandini Mullaji, told me that she’s always had a knack for setting people up. It runs in her family: Her grandmother is a matchmaker in India. But hiring a human matchmaker in the United States typically costs at least $5,000, Mullaji said. Sitch aims to provide a similar service—the ability to “really understand what your values, your preferences (both stated and not stated) are” and to find someone who suits you—for far less money.

Amata’s AI, Huraux told me, “learns about you, about your preferences; it curates profiles; it organizes the dates; it debriefs the dates.” In December, Justin McLeod, who founded the popular app Hinge, announced that he would be leaving the company to start his own AI matchmaking app, called Overtone. The traditional apps have already quietly incorporated AI into their existing platforms. Hinge users can sit back and let AI draft messages to their matches. And Bumble is rolling out an AI dating assistant called “Bee.”

As one of the many burned-out daters, I was tempted by the idea of a matchmaker, though the truth is that I mostly wanted to meet that tall man in the video. So I signed up for Amata to give it a shot.

My best friend, Michael, agreed to try Amata too. His AI matchmaker asked him at least an hour’s worth of questions about his past relationships, what he aspired to achieve in his career, what he was looking for in a partner, and his dating goals. Mine asked me far fewer questions: my height, my job title, my religion, one quality I was looking for in a man, and whether I was open to having children one day. Simple: 5 foot 7, journalist, Christian, wit, and probably. The back-and-forth lasted about two minutes.

After that, the process was relatively straightforward. My matchmaker avatar—a smiling red-haired woman—sent a few photos of a man that the AI had determined I might be interested in, with a paragraph describing the man’s dreams, how he spends his days, and what he’s looking for in a partner. My first thought was that my matchmaker didn’t know half as many of those details about me. What was it telling these men anyway?

The answer seemed to be that I would join a nunnery if I weren’t so determined to repopulate the world with babies. The men the app presented to me were heavily involved in their churches, volunteering multiple times a week. “He meets your must-have requirement of being a Christian,” the app kept telling me. Clearly, my matchmaker had misunderstood.

I tried correcting it: “Actually, I said having a partner of the same faith would be nice, but I said it wasn’t a must-have.” I tried to explain that I am a Christian, but I’m not that kind of Christian. (Think: less pastor’s wife, more raised by a Southern Baptist family that prays before dinner.) But my matchmaker wasn’t having it. “I know you are a Christian. How do you practice your faith?” I felt like I was being interrogated by my mother.

One of the benefits of AI matchmaking was supposed to be that it could help users get around their own biases—the expectations that hold them back from meeting people who might actually be the right fit for them. Many people, Huraux told me, have the “wrong criteria in their mind”: a strict height or age range that might make them miss a good match. But I didn’t suddenly become less shallow on Amata. My matchmaker showed me some men who sounded like they might suit me, but seeing them only two-dimensionally, I still found myself saying no before I could even finish reading the blurb.

I wasn’t the only one for whom the app malfunctioned. Allison Green, a 25-year-old woman living in New York, started using Amata after quitting Hinge. “I like it because it takes out the small talk,” she told me in January. But that was before she went on her first date. When I asked how it went, she said that the app had made a big mistake. Unlike me, she actually had told the matchmaker that she had a religious requirement: She wanted to date someone Jewish. And her date wasn’t Jewish.

She and the guy laughed it off, and set each other up with more suitable friends of theirs. “I don’t know if I trust Amata, though,” she told me. She kept feeling that the descriptions of the men weren’t accurate, and stopped using Amata a few weeks later. Whereas a human matchmaker knows if a client is lying about his height, and can at least guess if he’s lying about his personality, an AI matchmaker probably can’t (at least not yet). Much like traditional dating apps, AI matchmakers have to trust their users to be honest.

For my part, I now knew that the men in the Instagram videos who had drawn me to Amata were in a different category than the men my AI matchmaker thought I deserved. But I was having a hard time explaining that I wasn’t interested in the options it was offering me because they weren’t particularly … attractive. I asked Michael how he was navigating this. “Oh, I lost all political correctness within the first five minutes,” he said. He simply asked the AI, “Can you show me hotter people?” Meanwhile, I was replying “No, thank you!” like I might hurt its feelings.

Amata tries to discourage last-minute cancellations by requiring users to send apology notes. But three of the six men Michael was scheduled to meet called the dates off anyway. One match sent him a terse “Sorry shifting priorities!” the morning of. But another guy was much more decent. He explained that he couldn’t make the date they had planned because he had gotten last-minute tickets to a Lady Gaga concert—but would Michael like to come to that instead? Michael didn’t feel a romantic connection, but at least the app had more than paid for itself.

I was about to quit too, when the app gave me a reason to stick around: a party.

Every few months, Amata throws an in-person matchmaking event. This one was held on a rainy Wednesday night in March, at a bar on the Lower East Side. I had RSVP’d almost a month prior, but by the day of the event I had yet to receive confirmation that I was on the list. I reached out to explain that I was a reporter and wanted to attend the event for a story. Louis Munos, one of the co-founders, replied quickly to say that I was welcome.

I was surprised by how nervous I felt as I waited in line. I told myself that these people couldn’t be that intimidating—they were at a singles event, for goodness’ sake! But moves that once would have been considered sure signs of desperation—going to these parties, hiring a matchmaker, even just creating an online dating profile—are now commonplace. When love is this hard to find, our collective pride has gone out the window.

Once inside, I asked if anyone could point me to Munos. An energetic man with a very French French accent greeted me. “So your first matchmaking session will start in about 10 minutes,” he told me. I panicked: I had come to interview attendees, not to woo anyone myself. But Munos told me that I needed to get the “full experience.” A quick scan of the room told me that these men were much hotter than the ones I’d seen on the app. I was mad at my matchmaker.

Determined to use my 10 minutes wisely, I made a beeline to the bar. Gin and tonic in hand, I surveyed the crowd for a pack of approachable women. I introduced myself to some who looked to be in their mid-to-late 20s. They told me that they had come as a group and weren’t big users of the app; they were mostly on it to get to the parties. Then our phones started lighting up. Time for boyfriend roulette.

The first girl got the most normal-looking guy I’d ever seen. I was second. I showed the group the photo of the man I’d been matched with. “Oh! I saw him,” one of the women said. “He’s actually super tall.”

We were supposed to find our matches among the masses. I messaged mine that I was standing at the entrance, and tried to force myself to make eye contact with the men who walked by. “Annie?!” I turned to find a tall, attractive man waving at me. All the anger I’d felt toward my AI matchmaker melted away; its sins had been atoned. His name was Alex, and he was 35 and a massive flirt. Also, he had a British accent. Soon after we started talking, I saw a photo of him appear on the projector screen behind us. “Isn’t that you?”

Alex, apparently, was an Amata-party regular. I asked if he’d gotten a date out of the previous party. Yes, he said—a tiny blond woman (not his typical type, he assured me). They’d gone out twice, but then she’d ghosted him.

Seconds later, a woman who fit that exact description grabbed his arm. “Alex? I was just talking about you,” she said. “Where was that awesome spot we went on our second date?” He would tell her if she ever texted him back, he joked.

This was entirely too good to be true. “Hi, I’m Annie Joy!” I interjected. She didn’t care. While Alex and this tiny heartbreaker chatted, her friend told me that they were headed to the bar. Before they left, she promised Alex she would text him.

“So that was her!” I said.

“Oh God, no,” he replied. “I met her at Five Guys.”

After 30 minutes, it was time to meet our next match. I opened my phone and tried to politely hide my disappointment: This second guy was no Alex. But this second guy also never replied when I asked how I should find him, so I searched the room for new companions.

I spotted a girl covered in sticky notes. Say hi I’m nervous, one read. Kiss me, said another. I respected her game. Her name was Tia, and she hadn’t received a first match at all. It was one of a number of technical issues that night: Some people never received matches; some matches left before their sessions began; others must have been hiding out from the person they were supposed to talk with in favor of chatting with someone they’d met organically in the room.

But the mood remained optimistic, and no one seemed too worried about the flaws—people kept telling me that they preferred to pick their own match anyway. It struck me that very few people were on their phone. Everyone was either engaged in conversation or scanning the room for potential suitors. Flirting was so much easier when you knew that everyone else was there to do the same thing.

Then Tia told me that she had just spotted her monthslong situationship walk through the door. They weren’t exclusive, but she was still surprised to see him there. The New York City dating scene continued to shrink over the course of the night.

[Amogh Dimri: The people who marry chatbots]

At one point, I saw, floating through the party like a local celebrity, the raspy-voiced woman herself, the one who had lured me in with her interviews of ridiculously hot people. Her name was Issa Santiago, and she told me that when she wasn’t modeling or creating content for Amata, she was studying for the LSAT. She asked me if I wanted to do one of the videos myself, and I declined, though I have never been more flattered.

When I later told Huraux about the glitches I’d encountered on my app and the snafus at the matchmaking party, he admitted that Amata had some kinks to work through. Everyone will have different experiences on the app, he said, because the AI is responding to multiple variables: That’s why my matchmaker had asked me so few questions compared with Michael’s, for example. But Huraux was confident that the technology would improve, and that people would come around: “I think that five years from now, the normal way to meet will be to have an AI matchmaker getting to know you very, very well.” He told me that as humans build more trust with AI, they’ll become more comfortable with it “taking care of your love life.”

But Tia and her friends were skeptical. They doubted their AI matchmakers’ ability to understand who they were and what they wanted. Like so many people I spoke with that night, they were on the app only to go to the parties, where they hoped to find a connection in real life. The party I went to was so popular, Huraux later told me, that the waitlist was 35,000 people long.

At the end of the night, one of my new friends, walking out with her arms draped around two men, shouted back at me: “This is exactly what we’re missing right now!” She wasn’t talking about her AI matchmaker.