why is online dating so hard now?

The honest answer is that it stopped being a search problem and became an attention market. You are not bad at it. The shape of the thing has changed.

This essay walks through what actually shifted. It is written for the person who is venting into a Google search bar at 1am, half wondering whether the apps got worse or they got worse. The apps got worse. Here is the structural reason, in the order it stacks.

the apps are not optimized for the outcome you want

You want to meet a person. The apps want you to keep opening the app. Those are different goals, and once you notice they are different, almost everything that feels off about modern online dating clicks into place.

A user who meets someone on Hinge in three weeks and deletes the app is a churned account. A user who swipes for nine months without ever meeting anyone is a loyal one. The product team behind the app has no honest way to tell those two stories apart at the metrics layer. Sessions per week looks identical. So the optimizer, which is the org doing what the dashboards reward it for doing, tunes the experience for whatever produces the loyal-account pattern. Variable rewards. Notification spikes when you are about to leave. New filters that promise the next batch will be different. A premium tier that ships an extra layer of dopamine on top.

This is not a conspiracy. It is just what happens when a company has to grow revenue every quarter and the only lever that ships fast enough is engagement. The intent gap between you and the product is not a glitch. It is the model.

That gap shows up as friction you cannot name. The match never replies. The conversation dies on the second message. The person who said they wanted something serious has been on the app for fourteen months. None of that is randomness. It is what a system looks like when the people inside it have a different success function than the people on it.

the swipe was a UX choice that quietly changed who shows up

Before the swipe, online dating was a text-heavy, profile-heavy thing. You read paragraphs. You wrote a message that referenced something specific. The throughput was low and the signal was high.

Tinder shipped the swipe in 2012, and within about three years every major dating product converged on a card-based, photo-first, binary decision flow. The thing nobody quite says out loud is that this changed the population of people the apps select for. A photo-first binary decision in under two seconds favors a particular distribution of faces. Everyone else fades into a stack the algorithm slowly stops surfacing. You can feel this from either side of the swipe. If you are conventionally photogenic, you get a lot of attention that turns out to mean very little. If you are not, you get almost nothing and conclude something is wrong with you, when actually the format is wrong with everyone.

The swipe also rewired what you do with a profile. You stopped reading. You learned to flick. The apps trained the muscle, and now you cannot use the apps the slow way even when you want to. The interface taught you a habit that defeats the thing you are trying to do.

you are talking to fewer people than the app implies

There is a phrase the apps use a lot: active users. It is structurally misleading. An active user is anyone who opened the app in the last 30 days. It is not anyone who actually answers messages or shows up to dates.

The internal funnel inside almost every dating app looks roughly like this. A large pool opens the app weekly. A smaller pool swipes meaningfully. A much smaller pool initiates conversations. A much smaller pool responds. A tiny fraction of that responds twice. The number of people in your local area who are actually willing to have a back-and-forth with someone new this week is on the order of dozens, not thousands.

That is not because nobody is looking. It is because the format burns people out fast. Most of the people who downloaded the app this month are already exhausted from it. They open it, scroll for ten minutes, and close it without sending anything. From the dashboard they look active. From your inbox they are functionally not there.

This is why "there are no good people on the apps" is both wrong and right at the same time. The good people exist. They are just not currently emitting any signal that the app can show you, because the app rewards the wrong shape of behavior for them to participate.

the matching layer has the wrong inputs

Every recommendation system is only as good as the features it can read. A song recommender can read the audio, the tags, the play count, and the skip rate. A dating recommender, in principle, should be reading values, conversation style, sense of humor, and how someone treats a stranger. In practice it reads the photo, the bio length, the swipe rate, and the geolocation.

This matters because the things that actually predict whether two humans will like each other in person have almost no presence in the inputs. Voice does not exist in the feature vector. The way you describe a thing you love does not exist. The way you handle being mildly disagreed with does not exist. So the system optimizes for what it can see, which is attractiveness as encoded by other swipers and proximity, and it gets the rest by luck.

If the matching layer cannot read the thing that matters, then no amount of "improving the algorithm" inside that paradigm helps. The algorithm is reading the wrong file. That is a feature problem, not a model problem, and you cannot fix a feature problem by tuning the weights.

the time cost is hidden but enormous

A conversation on a dating app, end to end, from match to first decent in-person date, regularly takes seven to fifteen hours of phone time per person. Most of those conversations do not lead anywhere. So the implicit cost of one real date, on average, is dozens of hours of low-grade attention spent on people who never become anyone in your life.

Compare that to almost any other way humans used to meet each other. A hobby that takes three hours a week and incidentally puts you in a room with the same fifteen people for six months is a much cheaper way to find a person, by total time spent, than the apps are. The apps feel low effort because each individual session is short. They are not low effort. They are sliced into small pieces so that you do not notice the total.

When people say online dating is hard now, part of what they mean is that the per-real-meeting cost has gone up. The apps still work the same way they did in 2014, but the conversion from match to anything has fallen, the noise has risen, and the time tax compounds.

the cultural layer caught up to the burnout

For about a decade, online dating had a quiet social premium. It was a thing people did and did not quite admit to. That dynamic broke around 2022 or 2023. Now the dominant cultural script around dating apps is fatigue. The same friends who used to recommend Hinge now recommend deleting Hinge. The Reddit threads are venting threads. The TikTok essays are autopsies.

This matters because it changes who stays on the apps. When the cultural temperature drops below freezing for a category, the people who keep using the category are not a random sample. They are the people for whom the apps still kind of work, or the people who have not yet found something else, or the people who treat it as a hobby in itself. The shape of the pool changes. The median match changes. You can feel this even if you cannot name it. The room is thinner than it used to be, and it is mostly thinner of the kind of person you came in looking for.

what would make it less hard

This essay does not promise that anything will fix this fast. The honest list of what would actually help looks something like this.

A read mechanic instead of a swipe. Anything that asks the user to read for thirty seconds before making a decision. The swipe rewires the brain inside three sessions; a read mechanic does not.

A matcher that ignores photos at the recommendation layer. Photos can show up later, once a real exchange has already happened, but not as the first input. The job of the matcher is to find people whose interior is compatible. Photos are how mass-market dating got broken; they cannot also be the fix.

A culture where conversation can be slow without being penalized. Most app conversations die because someone took six hours to respond and the other person assumed disinterest and moved on. The format trained that. A format that allowed letters-length exchanges, or held the conversation gently open, would change which people stay.

Fewer optimizers. A product that does not need to grow forever does not need to addict you to grow. There is a class of small honest products inside almost every commodified category, and dating is going to need its version of that.

This is the part where I have to disclose something, since the rest of this essay was an honest answer and the closing part is not exactly disinterested. I work on byvibration, which is an attempt at this last list. It is photo-blind at the matching layer. It is write-first. It is built around a slower mechanic. The matching engine is open source at github.com/donnowyu/soulmate-core so the read-blind claim can be inspected rather than trusted. You can try it at byvibration.com if any of this resonated.

But the larger point of this essay does not depend on you trying any specific app. It depends on noticing that the difficulty you are feeling is structural, not personal. If online dating feels harder than it used to, that is because it actually got harder. The apps changed. The pool changed. The matching layer is reading the wrong features. The culture caught up to the burnout. None of that is something you are doing wrong.

The good news is that if the difficulty is structural, the solution is also structural. Either the existing apps reshape themselves, which is unlikely, or smaller honest products grow up alongside them and slowly absorb the people who left. Probably the second. It will take a while. Until it does, the best move is to spend less time on the apps, more time in rooms with the same humans every week, and keep an eye out for products that are obviously not optimizing for your engagement.

You are not bad at this. The format is bad at the thing you came in trying to do.