AI matchmaking features, reviewed — what they actually do

“AI matchmaking” now appears in the launch copy of nearly every new dating app. The label covers at least three quite different features, with different amounts of evidence behind them and different implications for the user.

Separating them is most of the review.

The three things the label covers

The last row is the point. Profile and conversation assistance are visible: a user can see what was suggested and judge it. Matching and moderation happen where nobody outside the company can observe them, and claims about both are essentially unfalsifiable from a review desk.

What works

Profile assistance is the clearest win. The problem it addresses is real — many people find writing about themselves difficult, and a blank prompt goes unanswered. A tool that helps someone produce a serviceable answer raises the floor across an app, and better profiles benefit everyone browsing them.

Photo selection help is quietly useful. Choosing which photographs represent you is a task people are demonstrably poor at, and a feature that surfaces which images others respond to is straightforwardly informative.

Moderation is where the technology is most valuable, and it is the application least marketed. Automated detection of harassment, scam patterns and image reuse operates at a scale human review cannot, and where an app describes its approach specifically, that is a meaningful signal.

What is doubtful

Conversation assistance is a design problem, not a feature. A tool that writes your opener produces an exchange between two systems and two people who have not yet said anything to each other. The category’s stated purpose is meeting someone; automating the meeting part is at odds with it. Users also report becoming quite good at spotting generated messages, which makes the feature actively counterproductive.

Matching claims are unverifiable. “Our AI learns what you like” is a claim no user and no reviewer can test. Every dating app has always ranked and filtered; relabelling that as AI does not add information. Where a company describes its approach specifically — what signals it uses, what it optimises for — that is worth something. Where it does not, the claim is decoration.

Optimisation targets may not be yours. A ranking system is optimised for something, and engagement is the metric most readily available. An app optimising for time-in-app and one optimising for people leaving satisfied would build noticeably different systems, and almost nobody states which they are doing.

Authenticity is the standing cost. The more of a profile and a conversation that is generated, the less the interaction tells either person about the other. That is the category’s central risk from this technology and it is not addressed by any of the marketing.

Pros and cons

Verdict

Our position is that the label should be ignored and the features assessed individually. An app using this technology to help people write a decent profile and to catch harassment faster is using it well, and should say so specifically. An app claiming its matching is smarter has made a claim nobody can check, and it should be read accordingly — as a statement about its marketing rather than its product. What we would watch most closely is whether an app will say what its ranking optimises for, because that answer is the one that would actually distinguish the serious from the decorative.

Coverage here is for adults; the apps discussed require users to be 18 or over.