How it works
Two private sets of answers, put next to each other
Between Us is built around a single awkward fact: the honest answer and the answer you give in front of your partner are not always the same one. Everything below exists to make the honest one safe to give.
The sequence
Five steps, and you can stop at any of them
One of you creates a session
You make an account and start a couples session. That puts your own private run of the questionnaire in front of you straight away — you do not have to wait for your partner to be ready, or even to have agreed yet.
A session is just a container for two people's separate answers. Creating one shares nothing with anyone.
You send your partner a private link
When you want to invite them, the session issues a one-time link you can send however you like. They open it, read what joining actually means, and make their own account.
We store only a one-way hash of that link, which is why we can show it to you exactly once and why asking for another one replaces it. It expires after fourteen days.
You both answer, apart
44 short questions across 12 areas. Most ask how something is for you. A few ask how important it is to you. A handful ask what you think your partner would say.
Answers save as you go. You can see a count of how far the other person has got, and nothing else. There is no request either of you can make that returns the other's responses.
You each agree to compare
Nothing is compared until both of you have said yes, separately. The agreement is specific: it covers the comparison, and free-text answers only if you have additionally chosen to share those.
Either of you can withdraw it afterwards. Withdrawing does not notify or require the other person.
You read the results together
Where you align, where you genuinely differ, where one of you misread the other, three specific things worth discussing, and some exercises to try. In full, for free.
Both of you see the same comparison, written so that neither person is the subject and neither is the problem.
Consent
Agreement is a switch you can flip back
A lot of products treat consent as a checkbox you clear once on the way in. Here it is a state the session is in, and it can leave that state at any point, from either side.
Shared, once you both agree
- Category-level comparisons of your ratings
- Where your answers were close and where they were far apart
- Perception gaps, in both directions
- Discussion topics and exercises built from the pattern
Never shared
- Your individual answer to any single question
- Free text you have not explicitly chosen to share
- Anything from your individual tests
- Anything you write in a private reflection
The method
Your results are arithmetic, not an opinion
Every number on your results page is produced by fixed rules in code. Identical answers produce an identical page, today and next year. Nothing about the scoring is generated by a language model, because a model that gives a slightly different verdict each time is not a measurement.
Each rating question is compared as a distance. Questions about what you each prefer are treated differently from questions about how something feels: a difference in preference is just a difference, and the engine refuses to count it as a deficit.
Each comparison records the method version that produced it, so you can tell whether two results from different months were built the same way.
Why there is no compatibility percentage
We do show how many rating items you answered within a point of each other, because that is a plain count of a real thing. We will not turn it into a headline percentage and call it compatibility. That number would imply a validated instrument behind it, and there is not one — not here, and not on any site that shows you one.
Where the engine deliberately stops
It will not name a person as the cause of a difference, will not label anyone with a condition or a personality type, and will not recommend ending or saving a relationship. Those are not tone decisions. They are rules in the code that produces your page.
AI, specifically
Used for writing, never for deciding
Plenty of products are coy about this. Here is the whole boundary, in one place.
- Scoring any question, in any test
- Comparing your two sets of answers
- Deciding which perception gaps are notable
- Choosing your discussion topics or exercises
- Anything that decides what you are shown
- Longer written interpretation inside the optional paid reports
- The neutral summary in the AI Conflict Report, written from both accounts
- Suggested wording for a conversation you are dreading
Those sections are generated from the already-calculated results, not from your raw answers, and they are clearly marked wherever they appear. They are never generated without a person asking for them.
Your answers
What actually happens to the things you type
Where they live
In a Postgres database with row-level security, which means the rules about who can read a row are enforced by the database itself rather than by application code that could have a bug in it.
Who can read them
You. Our own staff access is restricted and audited, and we do not read individual answers to improve anything. Support can see that a session exists, not what is in it.
What leaves the system
When you ask for an AI-written report, the already-calculated results are sent to the model provider — not your names, not your email addresses, not your raw answers. Nothing is sent without you asking.
How to get rid of them
Delete your answers for one session, or your whole account, from your settings. No email to support, no waiting period, no dark pattern on the way out.
No advertising script loads on any page that shows your answers, your results, a report or anything you have written. That is a rule in the code, checked per request, not a preference in a settings panel. The privacy policy and the cookie policy spell out the rest.
Without a partner
Most of this works for one person
The individual tests need no account at all. You answer, you read the result, and nothing is stored against your name unless you are signed in and ask us to save it. They describe patterns in what you reported — they do not diagnose you, and they will never tell you what your partner is.
One limit worth repeating.
Some of these tests touch on manipulation, control and narcissistic traits, because people genuinely want help thinking about those things. A questionnaire cannot establish that somebody has a personality disorder, and ours does not try. It describes behaviour you have reported and points you toward someone qualified. It does not put a label on a person who never answered anything.
Stopping
Leaving is a feature, not a support request
You can stop a session at any point, from either side. The comparison is deleted. Each person’s own answers stay theirs until they choose to delete them. Your partner is not asked to approve it and we do not ask you why.
You can withdraw agreement to compare without stopping the session, if you would rather pause than end. You can export everything we hold about you, and you can delete your account outright, both from your settings.
This questionnaire describes what the two of you reported on one particular day. It is not a clinical assessment, it cannot predict whether a relationship will last, and it does not establish that either person is at fault for anything.