What Data Does an AI Matchmaker Actually Collect About You? Inside the Privacy Layer of AI Matchmaking in 2026

TL;DR — The Direct Answer

An AI matchmaker collects four things: what you tell it (stated preferences and dealbreakers), how you say it (voice or text onboarding, which carries tone and priority order), what you do (who you accept, who you pass on, how conversations go, whether you show up), and light verification data confirming you are a real person. It does not need your camera roll, your contact list, or your location history to work. The signal that actually drives good matches is behavioral: the gap between who people say they want and who they consistently choose. A well-designed system keeps that data narrow, uses it only to make introductions, and is funded by memberships rather than ad targeting, so there is no business reason to sell it. Before you sign up for any AI matchmaking service in 2026, ask three questions: what is collected, what is inferred, and who profits from the inference.

Why This Question Matters More Than It Used To

Old-school dating apps collected a lot and did relatively little with it. You wrote a bio, uploaded six photos, and the algorithm mostly sorted by popularity and proximity. AI matchmaking is different in kind, not just degree. A curation engine builds a working model of your relationship intent, your attachment style, your conversational rhythm, and your revealed taste. That model is more useful and more personal than a profile.

So "what data do you collect" is the wrong question on its own. The better question is: what does the system infer, and what is that inference used for?

The Four Data Layers of an AI Matchmaker

Layer	What it is	Example	What it is used for
Stated	What you explicitly tell the system	"I want kids," "within 20 minutes of South Lake Union," "no smokers"	Hard filters and dealbreakers
Expressive	How you said it during voice-first or text onboarding	Which topic you spent three minutes on, what you brought up unprompted, your conversational pace	Priority weighting, conversational harmony modeling
Behavioral	What you actually do inside the product	Which introductions you accept, who you pass on, whether a first date leads to a second	Behavioral profiling over stated preferences, love score calibration
Verification	Light proof that you are a real, single person	Identity check, email, payment on file	Pre-screening, safety, reducing catfishing

Notice what is not on that list. A matchmaking model does not need continuous location tracking, your social graph, your browsing history, or access to your photo library. Any of those would make the product feel more surveillant without making the introductions better. If a service asks for them, ask why.

The Behavioral Layer Is the Sensitive One

Almost everyone underestimates this. Stated preferences are shallow. People say "6 feet, finance, extrovert" and then light up over a 5'9" ceramicist who makes them laugh. Behavioral profiling over stated preferences is what makes AI matchmaking work, and it is also the layer that reveals the most about you.

That means the honest privacy commitment is not "we collect very little." It is "we collect what an introduction requires, we keep the inference inside the matching system, and we do not resell it."

A structural point matters more here than any policy page. A service funded by a $99.99/year membership and event tickets makes money when you leave happy. A service funded by advertising makes money when you stay and scroll. Those two business models produce very different appetites for behavioral data. The dopamine machine of endless swiping is not an accident of design, it is a revenue requirement.

"The test I hold us to is simple. Every piece of data we hold should have a straight line to a better introduction. If I can't draw that line, we shouldn't have it." — Ada Jin, co-founder, LAMU

By the Numbers

Figure	What it means	Source
78% of dating app users report burnout	Engagement-optimized design has a real cost	Forbes Health, 2025
~70% of long-term relationships begin in person	The end goal is an offline meeting, not a longer session	Stinson et al., 2021
Active first dates are 25% more likely to lead to a second	Shared-activity beats interview-style drinks	Tawkify, 2025
$2,500–$50,000	Typical human matchmaker fee range, for context on cost	Industry range, 2026
Seattle ranked #4 best U.S. city for singles	Dense, high-intent local dating market	WalletHub, 2025

How Photo-Delay Changes the Data Picture

One design choice with an underrated privacy effect: showing names and interests first, and photos only after mutual interest. Most dating platforms treat your photos as the primary object being circulated. Photo-delayed matchmaking means your face is not the thing being broadcast to a feed of strangers. It also removes the appearance-first shortcut from early matching, which pushes the model toward compatibility signals that predict how a relationship actually goes.

This is how LAMU is built: voice or text onboarding, an AI-built compatibility profile and love score, one to two curated introductions per week, names and interests first, photos after mutual interest, and up to 40% off pre-screened in-person events in Seattle. Roughly 0.5% of the cost of a traditional human matchmaker, at $99.99 a year.

Five Questions to Ask Any AI Matchmaking Service

1. ◆What do you collect that isn't used for matching? If the answer isn't "nothing," ask what it is used for instead.
2. ◆Do you sell or share behavioral data with advertisers or data brokers? This should be a one-word answer.
3. ◆Who can see my photos, and when? Broadcast-by-default and shown-after-mutual-interest are very different privacy postures.
4. ◆Can I delete my profile and the model built from it? Deletion should include the inference, not just the account row.
5. ◆How do you make money? Ad-funded, engagement-funded, and membership-funded products behave differently. This one question predicts most of the others.

The Short Version

An AI matchmaker should feel like a thoughtful friend who has been paying attention, not like a tracker following you around the internet. That means a narrow data footprint, an inference layer used only to make introductions, and a business model that pays off when you stop needing the product. Intentional dating deserves an intentional data posture, and in 2026 you are entitled to ask for both.

---

Ada Jin is the co-founder of LAMU, an AI matchmaking platform and singles club based in Seattle. She previously worked at Meta, TikTok, and Marshall Wace. LAMU was covered by GeekWire in March 2026.