Andrew, a 21 year old college senior, needed advice. Recently, he’d been spending more time alone with his friend, but he was beginning to develop romantic feelings for her. He felt he was in a state of limbo, and wanted to take the next step, but found it hard to push himself out of his comfort zone.
He tried asking his roommates for help, but they gave conflicting counsel, then proceeded to argue over who was right. Most of his friends are in fraternities and he says they were unlikely to offer tender, thoughtful guidance he was looking for. At a loss, Andrew, who is being referred to by his middle name out of concern for his employment, decided to try something new—asking AI for advice.
Meeno, the AI relationship coach that Andrew consulted, didn’t suggest anything groundbreaking. But just the process of chatting with the app, and reading advice on the screen that confirmed his own instinct to confess his feelings, gave Andrew the confidence he needed.
“I kind of use it as a mentor,” says Andrew. “When I need help, I can reach [for] it. It’s not going to be affecting all facets of my life. But when I do have a social situation that I feel uncomfortable or indecisive when trying to figure out what to do, I think Meeno would be a great way to solve that.”
Renate Nyborg, 38, says she founded Meeno in November 2022 to try and address the increase in loneliness among young people, particularly after the COVID-19 pandemic. Users can consult Meeno about any relationship—friends, colleagues, parents—not just romantic partners. “This generation was lacking even basic social skills, because they spent two years at home staring at screens not being able to pick up basic body language. The conclusion for me was really that this wasn’t really a clinical problem. This wasn’t depression. It was actually just not having learned to speak a language,” says Nybotg, who was previously the CEO of Tinder. “It’s a skill that you can learn, but if you haven’t had the chance to practice it, then you don’t know how to do it.”
There are dozens of AI assistants springing up with less lofty aims. Rizz, a dating assistant app, suggests responses for dating app users. (The slang word rizz is short for charisma, and the Oxford University Press, which named rizz as its word of the year, defines it as “style, charm, or attractiveness; the ability to attract a romantic or sexual partner.”) “A lot of girls and guys—guys in particular—just do not know how to communicate online. It’s awkward in the very beginning, especially coming up with the right opener. It’s time consuming, it’s like a second job,” says Rizz co-founder Roman Khaves. “And so they come to Rizz in order to help them relieve that barrier, that friction point.”
Artem Chernikov, a 35 year old who works in finance in Montreal, says he typically uses Rizz three or four times over the course of a dating app conversation. This might be after the person makes a witty remark that he needs to match, or if he’s looking for inspiration after a long day at work. “Often, I won’t even use what Rizz suggests, but it’ll inspire me and it’ll get my creative juices flowing,” he says.
Where previously Chernikov might have consulted a group chat with friends for suggestions, now he asks Rizz. Does he tell dates that he uses an AI dating assistant? “Of course not, no.” Khaves concedes this is common. “A lot of users are using AI discreetly,” he says.
Chernikov is confident that his personality still shines through, because he still sends most of the messages. “If I used Rizz for every text exchange—for every interaction—the conversation would sound really weird.”
Dating app incumbents are starting to incorporate generative AI into their products, too. Match Group, the company that owns Tinder, Match.com, and Hinge, released a feature on Tinder in 2021 that detects potentially offensive messages and prompts the sender to reconsider before they send and asks the receiver if they want to report the message. The company is also testing an AI-powered tool for Tinder that would help users select their best photos and a feature that explains why another user might be a good match.
But traditional online dating services are facing a challenge from new dating apps that are built with artificial intelligence at their heart—the AI matchmakers.
AI Matchmakers
The first online dating sites became popular in the mid-’90s. As user numbers swelled to the millions, the sites had to come up with ways to narrow down the dating pool. This was the origin of compatibility matching algorithms, says Liesel Sharabi, associate professor in the Hugh Downs School of Human Communication and Director of the Relationships and Technology Lab at Arizona State University.
To begin with, dating sites had users describe themselves and what they wanted in a partner, but they gradually found this approach didn’t work because people are not very good judges of what they’re actually attracted to, says Sharabi. More recently, apps have instead tried to judge compatibility based on people’s behavior—in other words, who are they swiping on?
These algorithms are more basic than many people realize—their primary function is to screen for things like gender and level of education—meaning dating apps mainly just offer a larger pool of potential partners, says Michael Rosenfeld, a professor of sociology at Stanford University who studies dating and the internet’s effects on society. “The extent to which the apps and the dating sites say they have algorithmic secret formulas that are really important or effective, I think that’s pretty well overstated.”
Sharabi believes AI could change this. “Looking at AI, there are all kinds of possibilities for helping to make the decision fatigue a little bit less pronounced when you’re just having to swipe through so many profiles, and also helping to make the matchmaking process better by just improving upon some of the algorithms that they’ve been using,” she says.
Studies on the various factors that tend to produce longer, happier relationships have revealed a few consistent patterns. For example, couples with more similar personalities tend to be more compatible. Aware of this research, Viktoryia and Yanina Strylets, founded SciMatch, a dating app that gives users a compatibility score based on their personalities. Rather than using a questionnaire to assess personality, SciMatch has trained an AI model to predict personality from people’s faces. The founders claim their technology can predict measures of personality from facial scans with 87% accuracy. (Some experts argue that personality traits can’t be predicted from facial attributes). Facial scans are quicker than questionnaires, and they also have the benefit of being harder to spoof, says Yanina. “We focus on taking personality traits from faces because faces don’t lie.”
If, on average, people tend to couple up with others who have similar personalities, and if people who look alike tend to have similar personalities, does this mean that we tend to be more compatible with people who look like us? Viktoryia thinks so, pointing out the Instagram page “Siblings or Dating?,” where users submit photos of similar-looking couples—and siblings that look like couples—as evidence for the theory.
Anirudh Mallick, a 32-year old software engineer who lives in Manhattan, has been dating someone he met on SciMatch for over four months. Their first date was the first he’d been on from SciMatch—he’d seen a few people with whom SciMatch gave compatibility scores in the 50s and 60s, but when his compatibility score with his current partner was rated at 87%, he decided to send a message. Mallick speculates that they might have been matched because their personalities are somewhat similar, but they also balance each other out.
Another app, Iris Dating, is taking things back to basics. All other considerations—distance, personality, politics—pale into insignificance when two people are sufficiently attracted to each other, says Igor Khalatian, founder and CEO of Iris Dating.
“The importance of attraction is backed up by our data. We see the percentage of likes being increased when we predict [attraction] very heavily compared to other factors,” says Khalatian. “It doesn’t matter if you don’t match on politics, or distance or other factors. If attraction is present, it’s a dominating factor.”
Iris Dating predicts attraction by having users register their like or dislike of a large number of photos, starting with random stock images and then narrowing down to pictures of people the algorithm predicts the user will find attractive. At the end of this process, Iris claims it will have trained an AI that understands your type. “You all of a sudden have a genie who understands your taste,” says Khalatian. “And now [that] genie can scan, let’s say 100,000 people, and automatically shrink it to a smaller database of people you find attractive.”
Khalatian believes that the attraction Iris claims to predict is a solid basis for longer-term companionship. “Although we don’t monitor that, I have a strong feeling that it leads to, at the minimum, a very passionate kind of relationship,” says Khalatian. “But I think it’s more long term because, just realize, it may never happen to you again in life.”
In a future that TV show Black Mirror has already imagined, AI clones could go on simulated dates to test for compatibility. One tech startup is working on a similar idea.
Snack, a Gen Z-focused dating app launched in 2020, has users create an avatar and train it to emulate them by talking with it. But bot-on-bot dates proved awkward. “We tried AI to AI and there just wasn’t the flirtation and the banter going back and forth,” says Kim Kaplan, Snack’s CEO and founder. For now, users can choose to talk with a potential date’s avatar to get a sense of whether they’re a good match before speaking with the real person.
Some experts remain unconvinced that AI will ever be able to predict something as messy and interactional as human attraction. “I’m skeptical of whether the algorithms will ever be able to predict in a useful way, who’s really going to like each other, so much better than random, which is where we’re at now,” says Rosenfeld.
But Sharabi thinks that if AI could start to put all of the various factors together—personality, attraction, chemistry—then it might start to be able to predict compatibility. “It’s when you start combining all of these clues that you really maybe are getting a little bit closer,” she says. “It’s really tricky, because there’s just so much that goes into it. But being able to draw on lots of different qualities of partners and potentially seeing how they’re going to interact with each other—using AI to actually have conversations for them. That’s taking this in a bit of a different direction, where I think maybe we could get somewhere.”
More Must-Reads From TIME
