GraceMeet uses AI to suggest matches. We think anyone trusting us with something as serious as the search for a spouse deserves to know exactly how that works, so this page sets it out plainly, including the parts that are unflattering.
Once a week, an AI reads your profile and the profiles of people who could plausibly suit you. It scores each possible pairing against a fixed set of criteria, weighted towards faith and shared values. If someone scores highly enough, we show them to you along with a written explanation of why, and an honest note on what might be difficult.
Nothing else happens automatically. You decide. If you both say yes, a conversation opens. If either of you says no, that is the end of it and neither of you is told anything further.
Matches are worked out for each person separately. When we show you someone, they have usually not been shown you: their suggestions that week were calculated from their profile, not yours. The AI does occasionally pick two people for each other in the same week, but that is uncommon.
If you say yes, your profile is then shown to them, and they are told that someone has said yes before they answer. They do not have to work it out; we say so plainly. The alternative is worse: without it, a yes can disappear into a queue the other person never sees, which is exactly what used to happen.
If you say no, they are told nothing at all. They never see your profile, and never learn there was anything to respond to.
This does mean the two answers are not given on equal terms. Whoever answers first does so without knowing; the second person answers knowing. We think that is the right trade for making sure a yes actually reaches the person it was meant for, but it is a real asymmetry and you should know about it before you decide. The AI is told as well, and is instructed to treat it strictly as a tie-breaker: it may never raise a compatibility score, and it can never put someone in front of you who would not otherwise have met the threshold. That check is applied after the scoring, not during it.
This is the distinction that matters most, so we want to be exact about it.
The AI suggests. It never decides. It cannot introduce two people. It cannot open a conversation. It cannot see your messages. It has no say in whether a relationship continues. Every consequential decision on GraceMeet is made by a person, and a conversation only ever begins because two people each chose it.
The AI is doing the work a thoughtful friend might do if they knew a hundred single people: reading carefully, noticing who might genuinely suit whom, and making an introduction with reasons attached. It is a tool for narrowing, not an oracle, and certainly not a matchmaker with authority over anyone's discernment.
Every possible pairing is scored on four dimensions. The weightings are fixed and identical for everyone:
| Dimension | Weight | What it covers |
|---|---|---|
| Faith alignment | 30% | Theological and denominational compatibility, shared practice, and how each of you sees spiritual leadership in marriage |
| Values congruence | 25% | Shared faith values, views on marriage, and whether you want the same things regarding children |
| Personality and communication | 25% | Communication styles, love languages, and the pace of life you each prefer |
| Life stage fit | 20% | Age compatibility, career and life stage, and location |
Within faith alignment, not everything counts equally, and we have deliberately made theology dominant:
We revised this deliberately. An earlier version of our matching scored two people highly because they both loved sacred music and attended weekly, while quietly under-weighting a real doctrinal gap between them. That is a pleasant-looking match and a poor foundation for a marriage. Now, where theological distance is large, the faith score is capped regardless of how much surface practice the two people share.
Some things override the scoring completely. A match will not be shown to you if:
We only show matches scoring 80 or above out of 100. That limit is enforced in our code, not merely requested of the AI, because we found that models will occasionally talk themselves into a weaker match when the pool is thin.
Receiving no matches in a given week is a normal and correct outcome. It means nobody currently in the community met the bar for you. It is not a fault, and it is not a judgement about you.
We would rather send you nothing than send you someone we do not believe in. GraceMeet is a young and growing community, so quiet weeks are common at present. As more people join, they will become rarer. Padding your matches would be easy, and it would make the product feel more alive, but it would waste your time and quietly corrode the only thing that makes any of this worth doing.
You will receive at most three matches in a batch. Matches stay open for fourteen days so that nobody is left waiting indefinitely.
The system is not static. Three things shape what you are shown over time:
None of these can override the threshold. They shape which good matches rise to the top; they cannot rescue a weak one.
Some of these are technically easy and would probably increase usage. We are not going to do them.
Things we would want a careful person to know before trusting this:
Matching currently uses Claude Sonnet 5, made by Anthropic. We are a customer of Anthropic, not the developer of the model. When we change model we say so here and in our privacy policy. Your profile is sent to the model to generate your matches; it is not used to train Anthropic's models.
You can choose whether your data is used to improve GraceMeet's own matching. That setting lives in your privacy settings and you can withdraw it at any time.
If something here concerns you, or you think our approach is mistaken, we would genuinely rather hear it. There is a feedback option in the app, and you can write to us at gracemeetapp@gmail.com.
Several things described on this page exist because members told us we had got something wrong. That is the most useful thing anyone has done for this product.