Bumble — Why the Dream of an AI Matchmaker May Be Doomed
Whitney Wolfe Herd expects AI to reshape romance. More than a decade after launching Bumble, she’s returned to the dating-app company with a renewed mission: to replace shallow swiping with an AI-driven, emotionally attuned matchmaking process. As she recently told The Wall Street Journal, she envisions “the world’s smartest and most emotionally intelligent matchmaker.”
Dating companies large and small are rushing to answer that call. Firms are already using AI for photo selection, bios and message drafting; the next step, executives say, is using machine learning to recommend prospects on the basis of deeper traits. Wolfe Herd has described plans for an app that asks people about themselves and employs a large language model to surface matches aligned on what she calls “the things that matter most: shared values, shared goals, shared life beliefs.” She’s reportedly working with psychologists and relationship counselors to train the system.
Startups and incumbents are moving in similar directions. New apps such as Sitch and Amata experiment with AI-driven questionnaires and bot interactions to present bespoke suitors. Meta recently announced an AI assistant for Facebook Dating plus a weekly “Meet Cute” surprise-match feature. At The Atlantic Festival, Spencer Rascoff, CEO of Match Group, said Tinder is piloting a one-at-a-time prospect presentation—“like a traditional matchmaker,” he told an audience—presented as a more thoughtful alternative to endless swiping.
But experts and historical evidence temper enthusiasm for algorithmic soulmates. Long before generative AI, many dating sites used questionnaires and compatibility scoring. OkCupid’s 2013 experiments found that telling people they were compatible mattered more than whether the computed score was truly high— and that profile pictures drove activity far more than profile text. As co‑founder Christian Rudder concluded in a blog post, “OkCupid doesn’t really know what it’s doing.” The platform’s experiments are chronicled in Quartz and detailed at Gwern.
Beyond limited historic gains, the fundamental challenge remains human chemistry. Decades of social-science research show no simple formula for what sparks attraction. Northwestern psychologist Eli Finkel has noted that serendipity and the unpredictable flow of conversation often determine whether a real-life spark ignites. In short, the only true test of chemistry is the human encounter itself.
That doesn’t mean psychological research is irrelevant. Attachment theory, for example, offers useful insights: secure attachment tends to function well across pairings, while anxious and avoidant tendencies can complicate dynamics. Amir Levine, co-author of Attached, has described how mismatched attachment styles can create instability. But Levine and others caution that such frameworks are tools for self-awareness—not deterministic pairing algorithms.
The promise of an AI that can do the emotional labor of courtship raises thorny questions. Wolfe Herd has suggested a future in which a “dating concierge” handles not just introductions but much of the courtship process. Critics fear that offloading emotional risk and improvisation to machines could hollow out essential aspects of human connection. Even proponents of AI matchmaking implicitly sell a tidy narrative—that a better algorithm equals a better partner—when in reality relationships depend as much on individuals’ growth and behavior as on initial matching.
For now, the most realistic role for AI may be pragmatic assistance—helping people write profiles, making introductions less noisy, or nudging users toward reflection—not replacing the uncertain alchemy of attraction. As the industry tests more sophisticated tools, one lesson is emerging: human unpredictability is not a bug to be fixed, but a central feature of romance. When a genuine connection forms, it often feels, as it always has, like something no algorithm alone could have predicted.
Reporting for this piece draws on a longer examination published in The Atlantic.
