What Is the Iris Dating App and How Does It Work?

Iris is a dating app that uses artificial intelligence to learn which faces you find attractive, then shows you people predicted to find you attractive in return. You train it by rating pictures, and you can only message someone once you both like each other. The claims of better odds on its site are the company's own, and it promises no result.

You start by rating stock photos. Each rating teaches Iris what you respond to. You can also upload a photo of someone you find attractive to give it a sharper example.

Iris says it looks at facial features and is blind to skin color, race, hairstyle and clothes. That is the company's description of its method, not something we have tested.

The pitch is simple. Instead of scrolling for hours, you tell the app which faces you like and let it do the sorting.

Iris says balanced ratings work best. Liking almost everyone dilutes your type, so the app learns nothing about you. Being too picky slows training, because it has too little to work with.

This matches advice that applies on every app. a dating coach says an app that sees a man like every woman shown has little to go on, and that selective liking gives better results. Iris is built around that idea, so honest, varied ratings are the way to get the most from it.

Rate as if you were choosing a woman to take out on Friday, not clicking for fun. Iris learns from honest answers, and a rushed session teaches it the wrong type.

Mix your ratings. If you like only one look, Iris has little to separate. If you like nearly everyone, it has the opposite problem. When Your Type later shows faces that look wrong to you, that is a sign the ratings were too lopsided, and by Iris's own advice the cure is a more balanced set.

Once Iris has learned from you, your results sit in three tabs.

Your Type shows the people you are predicted to like. Your Fans shows the people predicted to like you. Mutuals shows people who are both.

For a man the tab to watch is Your Fans. It is the closest thing to a list of people who may be interested in you. Mutuals is where the odds look best on paper, since both sides are predicted to want it. Your Type is a menu, and a menu does not mean she will say yes.

You can message only after both people like each other. That means no line goes to a woman who has not already liked you back.

Iris also has the Iris Flower. Sending one puts your page at the top for the woman, as a stronger sign of interest. Use it the way you would use a Rose on Hinge: sparingly, and only for a woman you actually want to meet. A Flower sent to everyone means nothing.

Every member verifies the profile photo with a selfie, and the minimum age is 18. That may cut down on fake accounts, though a dating coach's caution still applies: a verified face says nothing about her intentions or her honesty. A short call before any date is still the best check.

Iris's claims of better odds on its site are its own, and it promises no result. We have no independent evidence that it works better than the big apps.

What decides your result is mostly out of Iris's hands. It depends on your photos, since the app predicts from faces and yours is one of them. It also depends on your area, because an app is only useful where enough people use it.

If you live in a smaller city, an app with a smaller user base can leave you with a short list. Check how many people it shows you before you spend time on it.

Iris suits a man who is tired of scrolling and likes the idea of letting a system sort for him. It also suits a man with strong photos who wants his page in front of women predicted to like his look.

It does not suit a man who is hoping the AI will compensate for weak photos. The face-based prediction depends on the face you show it, and a dark or blurry photo gives it less to work with.

If you try it, treat it like any other app. Run it for two weeks, count the conversations and dates it gives you, and compare with where you were. If it gives you more, keep it. If not, move your time back to the app that works.