Why matchmaking is turning out to be the one thing agentic AI can't shortcut
There is a specific kind of person, increasingly common, who will let an AI model book their flights, choose their hotel, reroute them around a canceled connection, and negotiate — implicitly, algorithmically — with three different airlines before breakfast, and who will not, under any circumstances, let that same model pick who they marry. This is not a contradiction. It is, on inspection, one of the more rational instincts of the AI era.
Start with the travel data, because it is unambiguous in a way dating data rarely is. Phocuswright reported that 56% of U.S. leisure travelers used AI for at least one trip in the past year, up from 43% just nine months earlier — one of the fastest behavioral shifts the travel industry has tracked in a decade. Booking.com's own sentiment research put the number wanting to use AI for future trip planning at 89%. And critically, this isn't idle browsing: roughly 78% of people who've used AI for travel say they booked something based primarily on what it recommended. That's not a novelty metric. That's a category quietly getting rebuilt from the discovery layer down.
Now hold that up against matchmaking, where the AI story is going in almost the opposite direction. In December 2025, Justin McLeod — the man who built Hinge into the most successful dating app of the last decade — stepped down as its CEO to launch an AI-first matchmaking service, backed in part by Match Group itself, the company that owns the app he just left. It is, structurally, a founder betting against his own creation. And even he has been explicit in interviews that the job of AI here is to coach, not to automate — a distinction the industry has not always respected. Three Day Rule, a fifteen-year-old matchmaking service, launched an AI-powered version trained on input from sixty human matchmakers, and one of the first widely circulated reviews described matches opening conversations with identical AI-generated lines — two people's AI wingmen, essentially, dating each other, with the humans relegated to a supporting role in their own courtship. Meanwhile Barclays' 2026 research found that 56% of Gen Z daters now prioritize meeting in person over extended in-app conversation, specifically citing distrust of AI-mediated chat as the reason. That's a generation that grew up with the technology actively opting to route around it at the exact moment it matters most.
The honest explanation for the gap isn't that travel AI is smarter than dating AI. It's that travel is a data-rich, low-stakes optimization problem, and matchmaking is neither.
A flight itinerary has a few hundred meaningful variables — price, timing, layovers, seat class, loyalty status, weather risk — and every one of them is legible, quantifiable, and available in structured form from an airline's own API. Get it wrong, and the cost is a bad connection, and even that comes with a human customer service line as a backstop. Compare that to matchmaking, where the "variables" include a person's sense of humor, how they behave under stress, whether they're actually done with their last relationship or just say they are, and a dozen other things people are famously unreliable narrators about even to themselves. There is no API for self-awareness. AI-generated itineraries, for what it's worth, still get things wrong at a startling rate — one comprehensive review found nine out of ten AI-built trip plans contained at least one major factual error, from invented landmarks to logistically impossible routing. If that's the error rate on a domain with clean, structured, verifiable data, it's worth sitting with what the equivalent failure mode looks like on a domain with none of that.
This is also, not coincidentally, why AI travel tools are converging on the same design principle the good AI dating tools eventually land on too: assist, don't decide. The travel industry's own 2026 research shows trust in AI is highest for research and comparison and drops sharply the moment money and irreversible decisions enter the picture — which is why direct booking through a trusted brand or advisor still beats a chatbot for anything that actually costs something. Dating has the same shape, except the "cost" isn't a plane ticket. It's an evening, a story you'll tell people, possibly a great deal more than that. The stakes are exactly backwards from where the automation is currently strongest.
None of this is an argument that AI is useless in dating — coaching tools that help someone write a better opening message or notice a self-sabotaging pattern in their own profile are a legitimate, well-evidenced use case, and the apps doing that carefully are outperforming the ones trying to automate the relationship itself. But there's a difference between AI helping a person show up better and AI standing in for the judgment call — and increasingly, the data suggests people can feel that difference, even when they can't fully articulate it.
Which is the case, made from an unusually direct angle, for keeping a human being in that particular loop. Luvo Matchmaking, DoRelish's high-touch sister service, is built entirely on the premise the AI dating story is quietly proving out from the other direction: that introductions sourced by a person who has actually met both people — not scraped, not scored, not generated — are still the version of this that works when the thing being optimized for isn't logistics but judgment. AI will keep getting better at booking your flight. It's going to be a while before it's better at knowing you.