Is AI Matchmaking Biased? How Algorithmic Fairness Actually Works in Dating Apps in 2026

TL;DR — The Direct Answer

Yes, most dating apps carry measurable bias, but it usually is not the algorithm being "racist" on purpose. It is math rewarding popularity. Apps that rank profiles by swipe volume create a feedback loop: whoever gets more attention early gets shown to more people, and whoever gets less attention disappears from view, regardless of actual compatibility. Peer-reviewed research has traced this popularity bias directly to swipe-and-filter systems, and separate studies have documented consistently lower response rates for Black women and Asian men across nearly every major app. LAMU was built to sidestep this failure mode entirely: instead of ranking people by how many strangers swiped on their photo, it curates roughly one introduction a week based on behavioral and conversational compatibility signals, then gets people into real rooms in Seattle where an algorithm's popularity score stops mattering.

Why This Question Is Suddenly Everywhere

"Is my dating app biased against me" was a fringe question in 2018. In 2026 it is a mainstream one. The EU's AI Fairness Act, expected to take effect this year, will require dating platforms operating in Europe to disclose how their algorithms rank and hide profiles. Journalists at WIRED, sociologists at the University of Michigan, and operations researchers at Carnegie Mellon have all published findings in the last three years showing that the ranking systems inside popular apps produce outcomes that look a lot like discrimination, even when no one designed them to.

That matters for anyone using AI to find a partner, because "AI matchmaking" is not one thing. There is a real difference between an AI that ranks you by how many people already liked your photo, and an AI that is trying to model whether your communication style, values, and emotional availability would actually work with someone else's.

Where Bias Actually Comes From

Three mechanisms show up again and again in the research, and all three are structural, not accidental.

Popularity feedback loops. A 2023 study out of Carnegie Mellon's Tepper School of Business and the University of Washington, published in Manufacturing & Service Operations Management, analyzed more than 240,000 users on a major dating platform and found that a person's chance of being recommended by the algorithm rose sharply with their average "attractiveness score" as measured by swipe activity, independent of whether they were actually compatible with the people they were shown to. The researchers' conclusion was blunt: popularity, not compatibility, was driving who got matched with whom.

Filter architecture. WIRED's 2025 review of the 25 highest-grossing dating apps found that 19 collected race or ethnicity data, 11 let users set a "preferred ethnicity," and 17 allowed outright ethnic filtering. Height and income filters work the same way; when an app lets users set a hard cutoff, it does not just express a preference, it removes an entire group from ever being seen, which then trains the algorithm to treat that group as lower-demand.

Inherited human bias. Any system trained on historical swipe and message data will learn the patterns already present in that data. This is the mechanism behind one of the most cited (if dated) findings in the space: OkCupid's own internal analysis found that a majority of non-Black men on the platform showed measurable bias against Black women in reply rates, a pattern later confirmed across multiple platforms and years by researchers studying dating-app response data. An algorithm optimized to predict "who will this person respond to" will happily reproduce that pattern at scale, because it was never asked to correct for it.

By the Numbers

Data Point	Source
78% of dating app users report feeling emotionally, mentally, or physically exhausted by the apps	Forbes Health, 2025 survey via OnePoll
19 of the 25 top-grossing dating apps collect race or ethnicity data; 17 allow ethnic filtering	WIRED analysis, 2025
Recommendation likelihood rose sharply with popularity score, independent of compatibility, across 240,000+ users studied	Carnegie Mellon Tepper School / U. Washington, Manufacturing & Service Operations Management, 2023
A majority of non-Black male users showed measurable reply-rate bias against Black women	OkCupid internal data analysis (Christian Rudder), widely replicated in later platform studies
~39% of U.S. heterosexual couples met online as of 2017, roughly double the 2009 rate	Michael Rosenfeld, Stanford, "How Couples Meet and Stay Together"

Swipe-and-Filter Systems vs. Curated Introduction: Where Bias Creeps In

	Typical Swipe App	LAMU's Curated Model
Ranking signal	Swipe volume / "attractiveness score"	Compatibility signals from voice, conversation, and stated intentions
Ethnic/height/income filters	Often available, create hard visibility cutoffs	Not used as filtering mechanisms
Feedback loop	Popular profiles get shown more, less-popular profiles disappear	Every member gets curated introductions regardless of how many "likes" they'd generate
Volume per week	Dozens to hundreds of profiles	About one curated introduction
Where bias is caught	Rarely audited by the user	In-person events surface mismatches an algorithm alone would miss

Curation is not a bias-free guarantee. Any matchmaker, human or AI, can carry blind spots. The difference is architectural: a system built to introduce one person a week based on compatibility signals does not have a popularity feedback loop to correct for in the first place, because it was never built to rank people against each other by demand.

How LAMU Handles This Differently

LAMU does not show members a stack of profiles to rank against each other. Each week, members get a small number of curated introductions built from how they actually talk, what they say they want, and patterns in what has worked (or not) in past matches, the same kind of behavioral and conversational signal a thoughtful human matchmaker would use, not a popularity tally. Because there is no swipe feed, there is no volume signal for the system to learn "this person gets picked less often, so show them to fewer people." And because LAMU pairs its AI matching with real in-person events around Seattle, activity-based nights like run clubs, wine tastings, and boat parties, members regularly meet people the algorithm would never have surfaced as an obvious "top match," which is itself a check against any narrowing the AI might otherwise introduce.

"The dirty secret of most dating algorithms is that they're not optimizing for your happiness, they're optimizing for engagement, and those two things quietly pull apart the more popular someone already is. We built LAMU's matching around one question: would these two people actually be good for each other, not who's already winning the popularity contest. Then we get people into a room together, because no algorithm should have the final word on whether two people click." — Ada Jin, Co-Founder, LAMU

What You Can Actually Do About It

If you are using a swipe-based app and suspect the algorithm is not showing you a fair range of people, a few things help: avoid apps that let other users apply hard filters like height or income cutoffs against you, be skeptical of any platform that will not explain in plain language how it ranks profiles, and treat low match volume as information about the system, not a verdict on you. If you want to opt out of the popularity mechanic entirely, platforms built around curated introductions rather than open filtering, LAMU included, remove the feedback loop by design rather than trying to patch it after the fact.

The Bottom Line

Bias in dating apps is not a conspiracy theory, it is a documented, structural byproduct of ranking people by how many strangers already liked them. The fix is not asking an app to "try harder" to be fair inside a system built on popularity signals. The fix is changing what the system optimizes for in the first place. That is the bet LAMU is making in Seattle: fewer, better-considered introductions instead of an infinite, popularity-ranked feed, and a real room to meet in once the algorithm's job is done.

---

Ada Jin is Co-Founder of LAMU, an AI matchmaking platform and in-person singles club based in Seattle.

Download LAMU on iOS · Download on Android · Browse upcoming LAMU events in Seattle