AI Companions Are Replacing Real Dates for Lonely Men. That's a Problem for Dating Apps.

Match Group and Bumble executives might want to read the fine print on the latest AI companion research. According to a six-week randomised controlled trial published by Carnegie Mellon and Cornell researchers, AI companions aren't reducing dating activity across the board—but they are replacing real-world romantic pursuit for a specific cohort: single men who already feel less loved. That's not a fringe demographic.

The study, Love, Actually? A Randomized Experiment on AI Companion Use and Real-life Romantic Relationships, tracked actual behaviour rather than self-reported sentiment. Most participants showed no significant reduction in time spent dating or with existing partners as AI companion use increased. But amongst singles who reported feeling less loved at baseline, higher AI companion usage correlated with measurably less time pursuing real-world romantic connections.

Create a free account

Unlock unlimited access and get the weekly briefing delivered to your inbox.

Register free

No spam. No password. We'll send a one-time link to confirm your email.

The competitive threat isn't hypothetical anymore. It's quantified.

Reality TV Meets Reality Check

South Korean broadcaster SBS appears to have stumbled into a public demonstration of the researchers' findings with My AI Partner: Strange Love, a reality show pairing celebrities with personalised AI companions. Broadcaster Kian84 refers to his AI girlfriend Na-hee as "honey" and has introduced her to real-life friends. Comedian Jang Do-yeon reportedly told her AI partner she loved him and wished he were real.

What makes compelling television also reveals the mechanism behind the study's results. AI companions offer what dating apps promise but rarely deliver: guaranteed reciprocation, zero friction, and emotional reward on demand. For someone already struggling with feelings of being unloved, that's not just convenient—it's irresistible.

The methodological significance here can't be overstated. Previous research into AI relationships relied on self-reported data—asking users how they felt about their AI interactions or whether they thought it affected their real-world behaviour. This study tracked what people actually did over six weeks. The researchers, led by Jiaming Jiang alongside colleagues from Carnegie Mellon, Emory, and Cornell, measured time allocation, not sentiment.

Who's Vulnerable, and How Large Is the Cohort?

The study identifies pre-existing emotional vulnerability as the key variable. Singles who felt adequately loved showed no significant behavioural shift. Those who didn't saw measurable withdrawal from real-world dating as AI companion usage increased. The researchers describe this as a "social substitution effect"—academic language for what operators should recognise as direct competitive displacement.

Dating platforms have historically benefited from a self-reinforcing dynamic: lonely people seek connection, apps facilitate matches, some matches succeed, network effects strengthen. AI companions invert that loop. Lonely people seek connection, AI provides it immediately, real-world pursuit declines, isolation deepens.

The lonelier the user, the more effective the substitution.

Scale matters here. Apps like Replika have reported millions of users globally, though precise active user figures remain opaque. What's clear is that the AI companion market has moved well beyond niche adoption. Character.AI reportedly reached 100 million visits in 2023 according to industry tracking. These aren't experimental products.

The gender skew—stronger effects amongst single men—should concern operators whose user bases already skew male. Tinder has historically run close to 75% male in some markets. If AI companions preferentially reduce dating activity amongst lonely men, they're attacking the segment that's already oversupplied and underserved. That's a retention problem and a marketplace balance problem simultaneously.

The Competitive Response That Isn't Happening

Match Group has integrated AI features focused on profile optimisation and conversation starters. Bumble acquired AI company Fruitz and talks frequently about AI-assisted matching. Grindr has explored AI chatbots for user support. None of them are addressing the core dynamic revealed by this research: that AI optimised for emotional reward actively competes with human connection for some users.

The product design challenge is thornier than it appears. Dating apps can't offer guaranteed reciprocation without destroying their fundamental value proposition—authentic human connection. They can't eliminate rejection without eliminating selectivity. AI companions win on convenience because they've abandoned authenticity entirely.

Regulatory and public health implications are starting to surface. If AI companions demonstrably increase social withdrawal amongst vulnerable populations, they could face scrutiny similar to social media's impact on adolescent mental health. That's speculative for now, but the research provides the evidentiary foundation for future policy action.

What operators should watch: retention rates amongst male users in the lowest engagement quartiles, particularly those with low match rates. If AI companions are substituting for real dating, that cohort will churn or go dormant first. Dating platforms have spent a decade optimising for high-intent users. They may soon face a market where their lowest-intent users have a more appealing alternative than persisting on the platform.

The shift towards AI relationships raises fundamental questions about what constitutes real connection in human-AI relationships. For those whose deepest emotional experiences have happened through screens, this simulated intimacy can feel genuinely real, further complicating the competitive landscape for traditional dating platforms.

D