I built a couples compatibility test where you "can't" see each other's answers
[▶ Watch the Do We Match demo on YouTube](https://www.youtube.com/watch?v=uifExOb1blY)](https://img.youtube.com/vi/uifExOb1blY/hqdefault.jpg)
[▶ Watch the Do We Match demo on YouTube](https://www.youtube.com/watch?v=uifExOb1blY)
I built Do We Match around a simple problem: Most relationship quizzes break the moment both people are in the same room — you soften your answers because the other person is watching.
So the core constraint of this build was that neither partner receives a copy of the other's raw answers ,not in the report, and not in the live AI conversation, which only ever receives your own answers. The written analyses do read both sides, so I added a programmatic check that rejects any generated text reproducing a partner's wording.
Both people therefore answer the same questions separately. Once both have finished, Do We Match brings the two perspectives together into one shared relationship check-up: where they align, where the gap is larger than expected, and what may be worth discussing next.
Then Matchie, our AI relationship consultant, can take the conversation deeper through optional analyses and private deep-dives. Matchie uses relevant context to explain patterns, ask adaptive follow-up questions, and turn a vague difference into something a couple can actually discuss.
Privacy created the harder design problem. A partner receives the shared comparison, not a copy of the other person’s full submission. I also did not want Matchie to become a generic chatbot or a breakup oracle. Its role is narrower: help two people find the conversation underneath the gap.
Product: https://dowematch.org/
Methodology: https://dowematch.org/methodology
What would an AI relationship product need to show or explain before you would trust it enough to use with a partner?
The privacy constraint is what makes this particularly interesting.
You’re asking people to trust the product with information they deliberately don’t want fully exposed even to the person they’re using it with. That’s a very specific kind of trust to earn.
Exactly! Thaat’s probably the hardest and most important part of the product.
The value only exists if both people feel safe enough to answer honestly, without worrying that every raw answer will be exposed to the other person. So I’m treating privacy as part of the product itself, not just a policy page.
The goal is to reveal useful patterns:Mismatches,shared values, and topics worth discussing, without turning the experience into “here is everything your partner said about you.”
I’m also thinking carefully about what gets stored, what each person can see and how easily users can delete their data. Trust will probably be the biggest factor in whether people actually use it honestly.
That’s helpful context. I appreciate you explaining how you’re thinking about trust as part of the product itself.
I’d like to continue the conversation outside the thread. What’s the best email to reach you on?
support@dowematch.org reaches me , I'd rather keep the useful parts here where other people can read them.
On the trust question, the concrete version: the live AI conversation only ever receives your own answers, so it has nothing of your partner's to leak. The written analyses do read both sides, so those run through a check that rejects any output that reproduces a partner's wording. That's the honest boundary , not "nothing is ever shared," but "here's exactly what crosses and what doesn't."
Thanks! I’ve just sent it over.
Looking forward to hearing your thoughts whenever you have a chance.