AI signs on as matchmaker in dating apps: it filters partners, but knows nothing of trust

By Hindustan Times · July 24, 2026.

The online dating industry has been losing paying subscribers for years and is now trying a way out: replacing the endless swipe with an artificial intelligence interviewer. Wavelength, an app launched in April 2026, offers "Wave," an AI companion that chats with the user about musical tastes, ideal partner type and weekend plans before proposing profiles; for Urvashi Gormat, one of its users in Bengaluru, it has suggested more than 30 so far. Knot.Dating does something similar in a 15-minute call for which it charges from 50,000 rupees (about $580) and, according to its cofounder Jasveer Singh, analyzes not only what the user says but the tone and speed of their speech. The British platform Romeo & Juliet, soon arriving in India, uses two different AIs: one interviews the client, the other cross-references the data to propose matches. Bumble has named its assistant "Bee"; Tinder is testing "Chemistry" and "Learning Mode" in Australia and New Zealand; and the founder of Hinge is preparing a new app this year, Overtone, with the same logic.

The context explains the rush. As Hindustan Times notes, more and more young people are choosing to stay single —what the English-language press already calls a "relationship recession"— and the sector's vocabulary has filled with terms like "swipe fatigue" (the exhaustion of endlessly rejecting profiles) or "microcheating." Traditional apps, with their infinite-catalog logic, have run out of answers to that fatigue. The CEO of happn, Karima Ben Abdelmalek, sums it up precisely: "A decade ago the goal was to expand options beyond your circle; now people don't want a buffet of options, they want to choose among three or four people they can genuinely connect with." Within that framework, AI is not selling more candidates: it is selling fewer, and better filtered.

There is one data point that deserves separate attention: the sector's gender gap, historically skewed toward men, appears to be closing. Wavelength reports 35% female users versus 65% male, and Knot.Dating reverses the usual ratio with 60% women. If confirmed over time, it would be a sign that taking the burden of the first screening —and of the initial exposure to harassment or rejection— off the user reduces one of the barriers that most kept women away from these platforms.

But the report itself offers the counterpoint, and it is the most relevant point in the text. The Bengaluru matchmaker Radhika Mohta notes that no client has ever asked her for "a six-foot guy with blue eyes for an ideal Sunday"; what people look for is someone who follows through on what they say, who doesn't bail on plans, someone who can be trusted. A 15-minute questionnaire, however well designed, does not measure that. Kaushik Banerjee, cofounder of Flutrr, himself admits that his bots connect the obvious —hikers with hikers, extroverts with extroverts— because that is the safe combination, not necessarily the one that sparks something. And no AI is trained to detect violent traits, a hidden marriage or a pattern of manipulation; that still requires time and human contact, the two things these apps promise to save.

Our reading is that this episode, though domestic, fits a pattern we have already seen in sectors with more at stake: AI takes over the administrative and repetitive part —the initial screening, the standard questions, the opening small talk— and leaves the final judgment intact, the one that requires experience and direct contact. We saw it in the legal profession, where AI empties the base of the pyramid but does not replace the senior partner's judgment; we see it here, where it filters profiles but cannot vouch for anyone's trustworthiness. It is a real automation, probably useful for anyone who hates the awkward first message, but it does not solve the underlying problem the article itself points out: all the apps fish in the same pond of people, and no AI can invent someone who isn't there.

In the short term it is worth scrutinizing the business model closely: an AI interview that costs $580 is not a democratizing service, it is a premium subscription with a barrier to entry, and the voice-tone analysis that Knot.Dating touts raises privacy questions the sector is only just beginning to consider. In the long term, if this automation delivers on what it promises —freeing people from the driest and most draining part of courtship so they can devote their energy to building something real—, it fits a broader thesis we have been defending in other areas: AI performs best when it absorbs administrative friction and leaves the human bond to humans. Love, for now, remains terrain where that bond is not outsourced.