<!-- What Is a Matching Algorithm? Dating Business Glossary | https://whitelabeldating.com/glossary/matching-algorithm/ | Run by High Intent Media, which also runs Dating Partners and DatingAPI. -->

# What is a matching algorithm?

> A matching algorithm is the set of rules a dating site or app uses to decide which profiles to show each member, and in what order. It weighs stated preferences, distance, activity and how members respond to each other.

## How does a matching algorithm work?

It filters the member base down to suitable profiles, then ranks them. Most dating platforms combine several kinds of signal:

| Type of signal | Examples | What it does |
|---|---|---|
| Stated preferences | Age range, gender sought, distance, relationship goal | Filters out profiles the member does not want |
| Compatibility | Questionnaire answers, shared interests, faith or lifestyle | Ranks profiles by how well two members fit |
| Behaviour | Who a member likes, messages and replies to | Learns what each member actually responds to |
| Activity | Last active, reply rate, profile completeness | Favours members likely to respond |
| Two-way interest | Whether the other member is likely to be interested back | Reduces one-sided likes that go nowhere |

Distance is usually handled by [geolocation matching](/glossary/geolocation-matching/), and the result shows up in your [match rate](/glossary/match-rate/).

## Why does a matching algorithm matter for a dating site or app?

It decides whether members find people worth talking to, which drives replies, paid upgrades and retention. A good algorithm also spreads attention: if a small number of profiles get most of the likes, many members see little interest and leave. Ranking that favours two-way interest and active members tends to give more members a conversation.

The algorithm cannot make up for an empty pool. If there are too few members nearby, no ranking will help; see [marketplace liquidity](/glossary/marketplace-liquidity/). That is one reason shared member networks exist.

## What rules apply to matching algorithms?

In the EU, a matching algorithm that suggests profiles is a recommender system under the [Digital Services Act](/glossary/digital-services-act/). Article 27 requires online platforms that use recommender systems to set out in their terms and conditions, in plain language, the main parameters used and any options members have to change or influence them.

Data protection rules apply as well. The ICO says inferred data counts as special category data where you infer a special category, regardless of how confident you are that the inference is correct. A matching model that infers sexual orientation from behaviour is therefore handling [special category data](/glossary/special-category-data/), which under GDPR Article 9 needs a condition such as explicit consent.

## What should you ask a provider about its matching algorithm?

Ask what signals it uses, how much you can adjust for your brand, and how it explains matching to members. Also ask:

1. Do members on your brand see members from other brands, and how is that decided?
2. Can niche brands weight faith, age or lifestyle more heavily?
3. Where are the main parameters explained to members?
4. Does the model infer any sensitive traits?

Our guide to [dating search and matching algorithms](/guides/dating-search-matching-algorithms/) covers the design choices in depth. This page is general information, not legal advice.

## Common questions

### Is a more complex matching algorithm always better?

No. On a smaller brand, having enough active members nearby matters more than clever ranking. A simple filter on age, distance and preferences, sorted by recent activity, often works well until the member base grows.

### Do members have to be told how matching works?

On online platforms covered by the EU Digital Services Act, yes: Article 27 requires the main parameters of recommender systems to be set out in plain language in the terms and conditions, with any options members have to change them.

### Can a matching algorithm use AI?

Yes, and many do. The same data protection rules apply, and if the model infers sensitive traits such as sexual orientation, that inference is treated as special category data.
