How Do Dating App Algorithms Work?

Entry No. 010 · October 2026

By The IF Team

Dating app algorithms are ranking systems. From everyone who fits your basic settings — who you want to meet, how far away, what age — they decide who you see first, and who sees you. Most weigh a similar handful of signals: how well two people fit what each says they want, how each has responded to profiles like the other’s, how recently each has been active, how complete a profile is and, on most big apps, whether someone has paid for visibility. The order is recalculated constantly as you like and pass. No major app publishes its formula, and much of what circulates online is guesswork presented as fact.

If your matches have dried up and you suspect you are being hidden, you are not paranoid: the order of a deck really does shape who meets whom. But you are probably not being punished either. So this guide labels its claims as it goes. “Said” means the company has stated it publicly. “Reported” means journalists or users describe it and the company has not confirmed it. “Speculation” means nobody outside the company knows, and a “myth” is a claim the evidence contradicts. Near the end, we explain the one part of IF’s ranking we make public, and the one thing no ranking can do: reach the person you already have in mind. That is what IF is for. We built it as a dating app for singles and people in relationships alike, and no conversation opens on it while either of you is in a relationship.

An algorithm filters first, then ranks

Every app does two jobs before you see a single face. The first is filtering: removing people who fall outside your hard settings, such as the gender you have said you want to meet, and anyone who has blocked you or whom you have blocked. The second is ranking: putting everyone left into an order. The ranking is where the interesting decisions live, because attention runs out. You meet the first profiles of an evening fresh and generous, and the fortieth with a good deal less patience.

A dating app also has a stranger job than a shop. A shop can recommend a jacket without asking whether the jacket likes you back. A dating app has to guess in both directions at once, so a good recommendation is someone you are likely to like who is also likely to like you. Researchers call this reciprocal recommendation, and it explains much of what feels mysterious: an app can hold back someone you would adore if it predicts they would pass on you.

The signals most apps use

The exact weights are secret, but the ingredients are not. From what the companies have said publicly, and from how recommendation systems generally work, these are the signals that matter most:

• What you say you want. Age range, distance and, on apps that offer them, filters such as height, religion or whether someone wants children. Some apps treat these as hard limits; others use them only to reorder.
• What you actually do. Who you like, who you pass on and who likes you. Apps lean on behaviour because stated preferences are a weak guide: in one speed-dating study, the ideals people described beforehand failed to predict whom they actually desired at the event.
• How active you are. A profile whose owner has not opened the app in weeks is a poor recommendation, because they may never reply. Tinder has said activity matters, and it would be surprising if any app ignored it.
• Whether conversations happen. Replies, and matches that turn into real chats, are strong evidence that a pairing works.
• How other people respond to you. Most recommendation systems use some measure of how people react to a profile. That is the kernel of truth inside the Elo story below.
• How complete your profile is. More photos and written answers give everyone more to go on.
• Money. Most large apps sell extra visibility.

The Tinder Elo score: what is true and what is myth

The Elo system comes from chess. Devised by the physicist Arpad Elo, it gives each player a rating that rises when they win and falls when they lose, and beating a strong player moves it further than beating a weak one. Applied to dating, a like works like a win: being liked by people who are themselves liked a lot raises your score more than being liked by people who are not.

In 2016, Tinder let a Fast Company reporter see his own internal rating, which staff called an “Elo score”, and the idea escaped into folklore, where it still lives. In a 2019 post, Tinder called Elo old news and said its technology no longer relied on it. The same post put activity near the centre: Tinder prioritises people who are active, and active at the same time. At its first product keynote, in March 2026, it presented an AI layer called Chemistry, which it describes as curating daily match recommendations from your interests, behaviour and personality signals.

So what survives? Speculation, though plausible: the way other people respond to your profile probably still counts somewhere in the system, whatever it is called now, because it counts in almost every recommendation system. Myth: that you carry one fixed attractiveness score you could look up, game, or reset by starting a new account.

The Hinge algorithm: what Hinge has said

Hinge has been more open than most. When Hinge launched Most Compatible in 2018 — a daily suggestion of someone it thinks you are likely to like and be liked by — it said the feature combined machine learning with the Gale–Shapley algorithm. That is a method for producing stable pairings, devised by the mathematicians David Gale and Lloyd Shapley in 1962. In plain terms, it tries to pair people so that no two of them would both rather be with each other than with the person they were matched to.

Hinge has also said Most Compatible learns from the profiles you like and skip, and the rest of the app very likely learns the same way. Because you like a specific photo or prompt rather than a whole person, and free members get a limited number of likes each day, each like is probably a clearer signal than a swipe. As of September 2026, its Dealbreakers work as hard filters: mark a preference as one, and people who do not fit it are not shown to you. Other preferences are treated more flexibly.

Is there a Hinge shadowban?

A shadowban means being quietly hidden while the app still lets you use it. Hinge has not confirmed doing anything it calls a shadowban, and nobody outside the company can prove it either way, so treat stories of one as speculation. What people describe usually has duller explanations:

• The novelty wore off. Apps are widely reported, though not confirmed, to show new profiles more in their first days. If that is true, a drop after a busy first week is extra exposure fading, not a penalty.
• Your filters are working. Strict Dealbreakers on your side, and on theirs, shrink the pool of people you can be shown to.
• You have been away. An inactive profile sinks on almost every app.
• The pool is smaller than it looks. In a small town, or with a narrow age range, most people nearby may already have seen you.
• A report is being looked at. Apps act on reports, and serious or repeated ones can get a profile restricted or removed. The app’s help centre is the place to ask.

If your likes have gone quiet, change one thing at a time — a new first photo, a looser filter, a week of steady use — and watch what moves. Our guide to why you might not be getting matches works through the likely causes in order.

The Bumble algorithm, before and after the swipe

Bumble has published less about how it orders its swipe deck than Tinder or Hinge. It is reasonable to guess that the general signals above apply — your settings, your activity, how people respond to you — but any detailed claim about it is speculation.

As of October 2026, what Bumble has said is mostly about where it is going. In May 2026, its founder and chief executive, Whitney Wolfe Herd, announced that the swipe would go in favour of AI-driven matchmaking. In August, Bumble said platform and data work had pushed that new version to early 2027, so until it reaches you, the Bumble you open still swipes.

At its centre is an AI assistant called Bee, revealed in March 2026, which Bumble says learns your values, relationship goals and communication style through private conversation and, when it finds two people it considers compatible, tells both of them why.

That is a real shift, and Tinder’s Chemistry points the same way. A deck algorithm reorders a queue you still choose from. A matchmaking algorithm chooses for you, a few people at a time, which makes its judgements more consequential and far more visible. Our guide to AI dating apps and matchmakers looks at how these assistants work and where they fall short.

Does paying boost you on dating apps?

Yes, in the plain sense. As of September 2026, the big apps all sell visibility. Tinder’s Boost and Bumble’s Spotlight put your profile near the front of the queue in your area for a limited window, and some top subscription tiers promise that your likes reach people sooner. None of this is hidden; it is on the price list.

Three things are worth knowing before you pay. First, a ranking is zero-sum: when one profile moves up, every profile it passes moves down a place, so a boosted evening is busier for you and slightly quieter for others. Second, paying changes how many people see you, not how they respond. If your profile is not landing, a boost shows it to more people who pass. Third, the suspicion that apps deliberately bury free members to sell subscriptions is common, but we have found no published evidence of a hidden penalty. Free tiers have visible limits, such as a daily cap on likes, which is not the same thing. Our comparison of dating app prices sets out what each tier includes.

What no algorithm can predict

Here is the humbling part, for the apps and for anyone hoping to crack them. In a later study, some of the same psychologists gave speed daters questionnaires covering more than 100 traits and preferences, then sent them on a series of four-minute dates. Using machine learning, they could predict, modestly, how much each person tended to desire others and to be desired, but not which particular pairs would click. The spark between two specific people only appeared when they met.

That sets the real ceiling on any dating algorithm. It can put plausible people in front of you and keep obviously wrong ones away. It cannot know whether you will laugh at the same moment. So the most useful things you can do are the ones that give it, and the people it shows you, a fair chance:

1. Choose; do not sweep. Liking everyone teaches the system nothing about your taste.
2. Show up in short, regular visits. Activity matters almost everywhere, and a few quiet minutes most evenings tend to serve you better than an occasional binge.
3. Answer your matches. Replies are among the clearest signals that a pairing works.
4. Hold filters loosely. Keep hard limits for the things you genuinely cannot live with, and let everything else reorder rather than exclude.
5. Improve the profile before you pay for reach. Our online dating tips cover photos, bios and first messages.

What we publish about IF’s deck

Every section above has had to hedge, because no big app shows its working. The part of ours we publish is Needs and Wants. On paid plans you can rank your preferences and split them into two groups. A profile that misses one of your Needs sinks far down your Swipe deck, but it is never excluded outright, because people are more than their filters. Wants count by rank: the higher a Want sits in your list, the more it counts. Our guide to Needs and Wants helps you sort yours. You also decide whether you can be found at all: turning off Discoverable in Search, which is free, means anyone searching your email, phone number or handle sees nothing, as if you were not on IF. The rest of the order stays private, because a published formula is a gamed one.

When the spark has already happened

All of this is about strangers. But plenty of people opening a dating app tonight already know who makes them laugh at the same moment — the friend of a friend who always ends up beside you at parties, the person from your Sunday league who already has a partner, the one whose face you half hope will turn up next in your deck. No ranking helps there. The question is not whether a deck will show them to you, but whether to say something.

Feeling that way is normal and honest, even when it crosses relationship boundaries. The only lines are cheating and lying, and you need neither. You can tell them, but opening up face to face has risks: it puts them on the spot and changes things between you before the timing is right. You are potentially better off sending an IF, a private, sealed message that opens only once you are both single — no conversation ever opens on IF while either of you is in a relationship — so that if the timing ever lines up, you are already there to show them how much you have always liked them.

Where IF fits

IF is the dating app for everyone, single or in a relationship, built so that you can tell a specific person how you feel, whatever the timing. Find them by Search, with an email address, phone number or Instagram handle they have shared with you, and nobody is told you searched. If you are both single and they are on IF, send a Like; an accepted Like opens a chat. If either of you is with someone, or they are not on IF yet, send an IF.

An IF is a private, sealed message of up to 250 characters, and no conversation ever opens on IF while either of you is in a relationship. If you are single and they become single, you get 24 hours to reveal it; if you are the one in a relationship, it reveals when you change your own status to single, as long as they are single too. Once it is revealed, they see who you are and what you wrote and choose whether to accept, and accepting opens a chat with your IF as its first message. How an IF works covers every case.

If they are on IF, they may be told that an IF is waiting — never who sent it or what it says, though people can sometimes guess from context. If they are not, an IF sent to their email address triggers one short invitation email, while one sent to a phone number or Instagram handle sends nothing and simply waits in case they join. Knowing that, the choice is yours.

Likes, matching and chatting on IF are free, and so is your first IF; Membership shows what paid plans add. Writing an IF takes a little emotional bravery, and that is rather the point: you say it once, mean every word and get on with your life while it waits. Right person, wrong time? Wait for them on IF. If the timing ever lines up, it may lead you to where you were meant to be all along.

Questions people ask

Does Tinder still use an Elo score?

No, according to Tinder, which said in 2019 that Elo was old news and that it no longer relies on it. How other people respond to your profile plausibly still counts somewhere, as it does in most recommendation systems, but Tinder has not said how much.

Does swiping right on everyone hurt your chances?

Probably, though the details are private. Indiscriminate swiping is widely reported to count against you, and the logic holds: liking everyone tells the system nothing about your taste, and a pile of matches you never message looks like pairings that do not work. Like the people you would genuinely write to.

Does deleting a dating app reset the algorithm?

Deleting the app from your phone usually leaves your account, and everything the app has learned, exactly where it was. Deleting the account and starting again is folklore as a reset: no app has confirmed it works, and you lose your matches and conversations.

Can I see my own dating app score?

No major app shows members their ranking score, and IF does not either. What you can see is the result: who matches with you, who replies and how the conversations go.

Can a dating app show me one specific person?

Not on a swipe deck, which shows whoever its ranking puts next. On IF you can find a specific person by Search, with an email address, phone number or Instagram handle they have shared with you, and nobody is told you searched. If you are both single and they are on IF, send a Like; if either of you is with someone, or they are not on IF yet, an IF says it now and waits until you are both single.

An algorithm can decide who you see first. It cannot decide who you have always liked: say it in an IF, and let it wait for the day your timing lines up.