How dating app algorithms decide who you see
No dating app publishes its formula, but the big three have said a fair amount about what goes into it. Here is what is on the record, what is guesswork, and what you can actually influence.
Updated
Try random matching insteadOpens videomegle in Video mode · free · no sign-up
What every dating app algorithm does
However they are marketed, recommendation systems on dating apps work in roughly three steps.
- Filter. First the app narrows everyone down to the people who fit both sides' basic settings: location and distance, age range, gender preferences and any dealbreakers. Nobody outside those limits will be shown to you, however compatible.
- Rank. Then it orders that pool by how likely it thinks a match is, using signals such as activity, profile details and past behaviour.
- Learn. Finally it updates the ranking from what you and others do: who you like, who you pass on, who likes you back, and who replies.
The companies describe the inputs in general terms but not the weights, and they change their systems often. Anything you read claiming exact numbers, such as a precise score or a secret swipe ratio, is speculation.
Tinder: from Elo to activity and AI picks
Tinder used to rank people with an internal score inspired by the Elo rating system from chess, which its then chief executive described in 2016 as a measure of desirability. In 2019 Tinder said it had moved away from it, and its own explanation of how matching works now calls Elo an outdated measure that its technology no longer relies on.
What Tinder says it uses instead:
- Activity. Tinder calls regular use the most important part of its algorithm, and says it prioritises potential matches who are active, and active at the same time.
- Location and preferences. Your current location, and your age, distance and gender preferences, as the starting point.
- Profile details. Interests and lifestyle details you add, and anonymised cues from photos, to suggest profiles similar to ones you have liked.
- Likes and Nopes. How often your profile, and profiles in your area, are liked or passed.
That last point is worth being honest about: even without Elo, how other people respond to your profile still feeds into who you are shown to. Tinder also says its algorithm does not track social status, religion or ethnicity.
In 2026 Tinder added an AI layer called Chemistry, which gives you a daily curated selection based on a Q&A and optional features such as a scan of your camera roll. Tinder says it has expanded from Australia and New Zealand to the US and Canada; elsewhere, its help centre describes a daily Most Compatible pick instead. It also announced a Learning Mode that adapts recommendations in real time from your activity. These are rolling out market by market, so what you see depends on where you are.
Hinge: Most Compatible and the Gale-Shapley idea
When Hinge launched its Most Compatible feature in 2018, it said the feature was based on the Gale-Shapley algorithm, a method published in 1962 by mathematicians David Gale and Lloyd Shapley for the "stable marriage problem". The idea is to pair people so that no two people would both rather be with each other than with the partners they were given. Shapley later shared a Nobel prize in economics for work on stable matching, which is why Hinge calls it a Nobel-prize-winning algorithm. Hinge adapted the idea to a dating app; it learns your preferences from who you like and skip, and tries to predict who is likely to like you back, not just who you will like.
Today Hinge's help centre describes Most Compatible as a single daily recommendation at the top of Discover, based on mutual preferences, recent activity and shared patterns, and lasting 24 hours. In 2025 Hinge said it rebuilt its Discover recommendations with deep learning to predict mutual compatibility. Separately, Standouts is a daily feed of profiles getting a lot of attention, which you can only contact by sending a Rose.
Bumble: less on the record
Bumble has said the least of the three. It describes using matching algorithms to predict compatibility and show you people it thinks suit you, and its Discover section shows a daily selection of compatible people based on shared interests, dating intentions and communities. It has not published how its main swipe queue is ordered. Your filters, location and settings clearly limit who appears, but beyond that any detailed claim about Bumble's ranking is unverified.
Where paid features fit in
Paying does change what happens, and the apps say so openly. Boosts and Spotlights move your profile up for a period. Priority likes are shown sooner. HingeX includes an always-on boost while you are subscribed and pushes people who are more your type higher in your Discover feed. What paid features do not do is change the underlying assessment of how compatible you are with someone. We look at whether they are worth it in are paid dating app subscriptions worth it, and at a single short boost in how Tinder Boost works.
Myths, and what you can actually influence
- "There is a secret attractiveness score." Tinder says it no longer uses Elo, but likes and passes still matter, so being liked more helps. The fix is the same either way: better photos and a profile people want to respond to.
- "Swiping right on everyone games the system." No app has said this helps, and since they learn from your choices, liking everyone gives them poor information about what you want.
- "You get punished for being picky." There is no published evidence of that. Liking the people you genuinely like is exactly what the systems are designed to learn from.
- "Your visibility can be secretly cut." Apps do take documented actions against rule-breaking accounts; see what a Tinder shadowban really is.
The things you control are the ones the apps keep naming: use the app regularly, keep your profile complete and current, set preferences that are honest but not needlessly narrow, and reply to your matches. If matches are still thin, start with why you are not getting matches.
Random chat works the other way round. On videomegle there are no profiles, scores or rankings: you choose video, voice or text and are paired at random with whoever is waiting. Our guide to how random matching works explains it.
Sources
Frequently asked questions
Does Tinder still use an Elo score?
Tinder says no. It calls Elo an outdated measure and says it now uses a dynamic system based on activity, location, preferences, profile details and how people engage through Likes and Nopes.
Does Hinge really use the Gale-Shapley algorithm?
Hinge said its Most Compatible feature was based on Gale-Shapley when it launched in 2018, adapted for a dating app. In 2025 it said its Discover recommendations were rebuilt with deep learning. It has not published the full details.
Is the daily Most Compatible pick someone who already likes me?
Not necessarily. It is the app's prediction of who you are likely to connect with. You still decide whether to like them, and they still decide whether to match.
Can I reset the algorithm by deleting my account?
It is not a reliable fix, and on Tinder having more than one account or making a new one after a ban breaks the Terms. Improving your profile and using the app regularly is the approach the apps themselves point to.