Justin McLeod Introduces Swipeless AI Matchmaking App

The promise of algorithmic matchmaking predates current generative AI systems. Dazed notes that algorithm-led matchmaking dates to the 1960s, including Technical Automated Compatibility Testing, or TACT. The publication also notes a well-known limitation of similarity-based systems: TACT reportedly matched a man with his younger sister.

Paul Eastwick, a University of California, Davis psychology professor and author of *Bonded by Evolution*, told Dazed that machine-learning research has attempted to predict relationship compatibility for roughly a decade. His conclusion was more cautious: "You can predict who's popular, but you can't predict who's compatible."

That distinction is material for ML practitioners. Ranking systems can optimize measurable engagement proxies, such as responses, likes, or meeting rates, but long-term interpersonal compatibility is harder to define, label, and validate. In comparable consumer recommendation settings, sparse outcome data and subjective targets can limit claims that a model has identified a durable preference rather than a short-term behavioral correlation.

The Los Angeles Times reported in June that users frustrated with swipe fatigue were trying AI matchmakers including San Francisco-based Known. Its reporting also found that larger dating platforms were incorporating AI features and selling subscriptions framed around smarter matching and faster connections.

The same Los Angeles Times article documented concerns from users and experts that automated matchmaking could undermine authenticity and real-world chemistry. Those concerns align with Eastwick's warning that pre-meeting data may not reliably determine whether two people are compatible after they meet.

Overtone therefore enters an increasingly active product category, but the core technical question remains evaluation. Companies offering similar systems face a difficult measurement problem: proving that a smaller number of model-curated introductions improves relationship outcomes, rather than simply changing the interface through which users make choices.