Tokyo's AI Matchmaking Service Reports 265 Marriage Outcomes
Tokyo's government said 265 couples had reached the marriage outcome in its TOKYO Enmusubi matchmaking service as of June 30, 2026. The program reported about 36,000 applicants, 16,000 registered users and 760 couples in serious relationships; the figures document a public-sector AI deployment, but do not by themselves show that the matching algorithm caused those outcomes.
Tokyo's government reported that 265 couples had reached the marriage outcome through its TOKYO Enmusubi matchmaking service as of June 30, 2026. The AI-assisted program began accepting applications on September 20, 2024, and is intended for adults who live, work or study in Tokyo and are seriously seeking marriage.
The official figures describe a substantial funnel, although they are not an evaluation of the matching model:
| Program metric | Cumulative total through June 30 |
|---|---|
| Applications | About 36,000 people |
| Registered users | About 16,000 people |
| Couples in serious relationships | 760 |
| Couples reaching the marriage outcome | 265 |
Tokyo defines registration as completing an interview, document submission and other enrollment steps. It defines a serious relationship as a couple choosing to stop new introductions and other dating through the service. A marriage outcome is recorded when both partners intend to marry and ask to leave the program; that administrative measure should not be read as independent verification of every legal marriage registration.
What the AI does
The service uses a values assessment and stated partner preferences to recommend compatible people. Users can also search by their own criteria, so the published outcomes reflect the full service rather than AI recommendations alone. Enrollment requires identity and single-status documents, proof of income, a profile photo, a pledge and an online interview; registration costs 11,000 yen for two years.
J-CAST reported that the main user group is in its 30s and that men account for about 60% of registered users. It also reported that the system can make up to four AI recommendations per month, alongside user-directed requests.
What the numbers can and cannot show
For recommender-system teams, these figures are useful because Tokyo reports milestones beyond clicks or matches: completed enrollment, serious relationships and marriage intent. Those are closer to the service's real objective than short-term engagement.
They are not, however, a causal performance measure. Tokyo has not published a comparison group, the share of outcomes attributable specifically to AI recommendations, model-quality metrics or demographic fairness results. The gap between applications and registered users also includes document and interview requirements, so it cannot be treated simply as algorithmic drop-off. The outcome data show real-world use at meaningful scale; evaluating whether the AI improved matching would require attribution, cohort and fairness analysis that the published figures do not provide.
Key Points
- 1Tokyo reported 265 couples reaching the marriage outcome, 760 serious-relationship couples and about 16,000 registered users as of June 30, 2026.
- 2TOKYO Enmusubi combines values-based AI recommendations with user search, identity checks, single-status documents and an online interview.
- 3The funnel shows real-world adoption and outcome tracking, but Tokyo has not published causal attribution or demographic fairness results for the matching algorithm.
Scoring Rationale
The program provides unusual public funnel metrics for a real-world AI-assisted matching service, including relationship and marriage-intent outcomes. It is useful applied-AI evidence, but Tokyo has not published causal attribution, model-quality benchmarks or fairness analysis, limiting the technical significance.
Sources
Primary source and supporting public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
