Opinion
The AI in your dating app is masking your red flags
Hinge's AI now shapes who you see, what you say and what you send. It is washing out the warning signs you would otherwise spot, and by the time you meet in person, the app has bowed out.
Photo by Alexander Sinn on Unsplash
“Dating has always been hard, and it’s always been uncomfortable.” So said Hinge CEO Jackie Jantos on the BBC’s Big Boss Interview last week. Luckily for us, Hinge has enlisted the help of AI so can now polish your profile, write your opening lines, and stop you sending a message it doesn’t like. Jantos was asked on air whether the person on the other end should be told. Four times. She said no.
So the person you’re falling for has been edited, and you don’t get to know by how much.
Hinge is “designed to be deleted.” Its job is to get you off it and onto a good date. By Jantos’s own account, Hinge deploys AI in three places, and together they cover almost everything that passes between two people before they meet.
First, the algorithm decides who you see. That, she said, is machine learning at its core, and Hinge has run on it for years. Second, a set of tools shapes what you say. And then handy feedback reviews your profile answers and pushes you to say more: “I love watching movies” is weak, so the tool asks what kind, and out comes Amelie, the Saturday matinee, the detail that makes you legible to a stranger. Convo Starters suggests your opening lines. Third, a safety tool screens what you send. It reads a message before it goes, scans for anything untoward, and asks “Are you sure?” All of it, Jantos said, helps “someone tell more of who they are in their own words.”
Who you see, what you say, what you send. When Sean Farrington pressed on whether the person you’re matched with should know, Jantos wouldn’t concede it. “No,” she said, “because the person is using their own words to describe who they are. This is not AI writing or performing as you.”
What “authenticity” edits out
Each tool still removes signal. The profile you read has been edited to the app’s idea of quality; the message that reaches you has had its rough edges sanded off. For most daters that’s harmless tidying. For someone testing how much control they can take, it smooths the very cues a careful person uses to judge whether someone is safe. Coercive control never announces itself in one message. It builds across weeks, the warmth, then the isolation, then the demand, and the surface is exactly where it hides.
That pattern is not hypothetical. Lewis Rollins, a Metropolitan Police officer, met Dr Georgia Miller on Bumble in April 2022. Across the relationship he monitored her phone, turned on her in bursts of anger, and pressed her for money: a thousand pounds for a bet he insisted would pay off, more again for his rent. He was running the same routine on several women he had met on apps at the same time. A jury at Southwark Crown Court convicted him of coercive control and fraud. The behaviour was a single pattern, played across multiple relationships at once, and the only place it was ever read as one was a courtroom, long after the money and the harm were gone.
The missing layer
Every platform now invests in AI that verifies a face and filters a single offensive message. None reads the pattern of coercion across a whole conversation, the layer that would catch a campaign of control before it has a name. It doesn’t exist on any app, and the Online Safety Act is heading toward it.
That layer does exist. Iris is a signal-processing engine, deterministic and auditable, not generative AI. It reads a whole conversation rather than judging one message at a time, scores the markers of coercive and manipulative behaviour in a way a court could test, and surfaces them early, with the evidence attached. It replaces no one. It empowers and connects the people already doing the work: the support worker, the police officer, the fraud team.
Jantos is right that the goal is authenticity. But the bigger question, the one she waved away, is whether an industry varnishing the surface is also dulling the instincts that would make you walk away. Until something reads for the pattern, AI will take the edge off the ‘ick’ and eventually muffle the alarm.
Sources: BBC Big Boss Interview with Jackie Jantos, June 2026; R v Lewis Rollins, Southwark Crown Court, coercive control and fraud; Online Safety Act 2023.
← Back to Iris · caroline@trustiris.com
Confidential. (c) 2026 Iris Anticipa Ltd. All rights reserved.