Dating Apps Without a Desirability Score: How It Works
A desirability score is the hidden number many dating apps compute from your photo and swipe history, then use to decide whose feed you land in and how often. Users are rarely told it exists. A 2025 peer-reviewed study in JMIR Formative Research found that mechanisms like this — including outright algorithmic match throttling — disproportionately harm men's psychological well-being, one line in a growing body of research on ranking systems nobody outside the company can audit.
Picture the ordinary alternative: someone reading on a moving tram, window light sliding across the screen, and nothing about how they look being scored while they do it. That's not incidental — it's how Anketta's matching actually runs. Hard filters go first: age range, gender, who you want to be shown to, relationship intent, city. Then a preference model builds itself from the words and phrases you highlight in someone's writing. Nothing about your face, your photo, or your appearance touches that pipeline — not because photos don't exist here, but because they were never wired into the ranking to begin with.
A desirability score is a hidden number a dating app assigns to your profile — built mostly from your photo and how other users react to it — that decides whose feed you land in and how often. Anketta doesn't compute one. Matching runs on hard filters (age, gender, city, intent) and a preference model built from what you highlight in someone's writing, not what you look like.
The mechanism is old, and it isn't unique to any one company. A score like this typically blends how often people react positively to your photo, how quickly you get liked back, and sometimes how selective you are with your own likes — then feeds that number back into who sees you next. It's a closed loop: the system learns from itself, and a user has no way to see the number or contest it.
Write the manuscript that gets read before it gets rankedThe best-documented example is Tinder's old Elo score — a chess-style rating, first reported publicly in 2016, that ranked users by desirability and matched similarly-rated profiles together. Tinder retired the name in 2019 but kept ranking exposure by engagement and mutual interest, a pattern most swipe-first apps still run in some form.
Fairness means naming what the old system actually solved: an Elo-style score is a reasonable answer to a real problem — showing people profiles they're statistically likelier to match with, instead of a random feed nobody clicks on. The company's own explanation for retiring it, reported by Engadget, was blunt about the label if not the mechanism underneath it:
"Elo is old news at Tinder. It's an outdated measure and our cutting-edge technology no longer relies on it." — Tinder, quoted by Engadget (2019)
What replaced it still starts from the same raw material — likes, dismissals, replies — and those signals are downstream of a photo, because a photo is the only thing on screen when the decision gets made.
People judge attractiveness from a face in about one-tenth of a second — a finding from a 2006 psychology study that held up even when viewers were given unlimited time to look. That's the signal a photo-first feed has to work with at the moment of a swipe, so any ranking system trained on those swipes inherits the same shortcut.
Willis and Todorov's 2006 study tested exactly this: participants rated faces for attractiveness, trustworthiness, and competence after seeing them for a tenth of a second, and their snap judgments correlated closely with judgments made with no time limit at all. Extra viewing time barely moved the number. A swipe interface asks for a decision in roughly that window — so whatever the system learns from swipes is, structurally, learning from the same fast, appearance-led read.
Anketta's ranking runs on two layers: hard filters first (age range, gender, who you want to be shown to, relationship intent, city), then a preference model that grows from what you highlight — mark a phrase as "love this" and the algorithm surfaces more writing like it; cross something out and it surfaces less. Your photo touches neither layer.
None of that fast-read shortcut is available to Anketta's ranking, because the manuscript it reads is the whole point of the interaction, not a screen you glance past on the way to someone's photo. Concretely, three things decide who you're shown to, and a fourth thing never enters the equation at all:
- Hard filters, applied first. Age range, gender, who you want to be shown to, relationship intent, city — the same coarse sort most apps use, just without a score layered on top.
- A preference model built from your highlights. Mark a phrase as "love this" and the queue tilts toward more writing like it; cross something out and it tilts away. The model only knows what you've told it, one highlight at a time.
- Nothing about your photo. It isn't in the payload the feed reads, it isn't scored, and it doesn't move you up or down a single place.
| Score-based ranking (e.g., Tinder's old Elo) | Anketta | |
|---|---|---|
| What decides your exposure | A desirability/engagement number computed from your photo and activity | Hard filters, then a preference model built from what readers highlight |
| Where your photo fits in | Determines exposure from the first screen | Hidden until an active match — see below |
| How you'd try to "improve" it | Better photos, more swiping, more activity | Write something someone actually wants to underline |
| Is the mechanism disclosed | Historically opaque — Tinder's Elo went undisclosed until reporters investigated it | Described in this article — hard filters, then highlights |
This is the same logic behind text-first dating generally — when the input a system reads is prose instead of a photo, appearance simply isn't a variable it has access to. The only real lever left is to write something worth highlighting.
Start a manuscript only your words can rankPhotos exist on Anketta — up to three, all optional — but they sit in no payload before a match: not in the feed, not in the reader view, nowhere a ranking system could read them. A photo only appears once a match is active, and only to someone who has shared one back.
Anketta isn't a no-photo app. What changes is the order: a photo only renders inside an active match, and even then only for someone who's shared one back, so nobody's face is ever the first thing a stranger reacts to. New uploads also show blurred until two other members approve them, or an admin does. The difference from a photo-grid app isn't that the photo is missing — it's that it arrives after there's already a reason to care what it looks like, not before. It's the same instinct behind the 48-hour match window: decisions that used to happen in a tenth of a second get pushed to a point where there's more than a face to go on.
Removing a desirability score doesn't guarantee a "fairer" outcome by itself — no honest article can promise that. What it changes is what the system is allowed to react to: it can't reward a flattering photo angle or punish a plain one, because appearance was never wired into the input in the first place.
There's no honest claim here that a filter-and-highlight system is bias-free — any ranking has assumptions baked into it, Anketta's included. What's specific to a desirability score is the particular failure the JMIR paper names: a number, invisible to the person it's ranking, that can throttle their exposure for reasons they'll never see and can't contest. A system with no photo in its inputs at all can't run that exact failure — not because it's smarter, but because there's nothing about your face for it to score in the first place. Someone still has to open a manuscript and write something worth reading. The app just isn't deciding, before they've read a word, whether that someone gets the chance.
Unsure about writing? Try reading first.Does Anketta compute an attractiveness or desirability score?
No. Matching runs on hard filters — age, gender, who you want to see, relationship intent, city — plus a preference model built from the words and phrases you highlight in someone's writing. No number tied to appearance exists anywhere in that pipeline.
Will uploading a better photo get me shown to more people?
No. A photo can't improve your reach on Anketta, because photos aren't in the feed's payload before a match — there's nothing there for a ranking system to read or score in the first place. What moves you is writing, and what a reader chooses to highlight in it.
How is this different from Tinder's old Elo score?
Elo ranked users against each other by a computed desirability number and matched similar scores together. Anketta doesn't compute a number at all — filters narrow the pool, then a preference model built from highlights shapes what you see next, with no appearance input anywhere in the chain.
If there's no score, how does Anketta rank anyone?
Hard filters run first, then a preference model learns from what you mark as a "like" or a cross-out in other people's writing. It's still a ranking system — just one whose only inputs are demographic filters and text, never a computed appearance number.
Are photos hidden completely, or just delayed?
Delayed, not hidden. Up to three photos are allowed and optional. None of them appear before a match; once a match is active, a photo shows only to someone who's shared one back, and new uploads stay blurred until other members approve them.
Does removing appearance from ranking actually change match quality?
It changes what the system can react to, which isn't the same claim as "better matches." A ranking with no appearance input can't reward a flattering photo or punish a plain one, because looks were never wired into the input to begin with.