How to read compatibility test results and your Love Score

August 11, 2026 · 15 min read

Reading a compatibility test result means reading a number against the thing it was measured on, not against other people. In LoveScore that number is the Love Score: one figure from 0 to 100, computed for the two of you out of eight of the nine relationship dimensions, once both partners have answered 47 questions on their own devices. It is not a percentage and not a percentile. The scale has two fixed anchors: a couple who answer “it depends” to every question land exactly on 50, and 100 is the theoretical ceiling, where both partners sit at the top of every level scale and agree wherever agreement is the thing being measured. The number is produced by a deterministic algorithm — the same answers always give the same score — and the language model writes the explanations afterwards, over figures it has no way to move.

A compatibility score is never the whole of the results, though, and the rest of them are where the usable part sits. Next to that number the report carries four more blocks: your strengths as a couple, your friction points, an archetype for each partner, and an alignment index with six scenarios. What follows is a reading of all five — what each one measures, how each one gets misread, and, the most useful part, what none of them can tell you. Your results page is private and is never indexed, so this is the place where the numbers get explained.

What is a good compatibility score?

This is the question nearly everyone arrives with, and the honest answer is disappointing for about thirty seconds before it becomes useful. There is no threshold above which a couple is fine, on this scale or on anyone else's. There is no percentile either: LoveScore does not know how other couples scored, because it holds no normative sample and never collected one. A service that tells you that you are “ahead of 78% of couples” either has a database it is not showing you or has invented the comparison. We would be inventing it, so the report never makes it.

What the scale does support is a comparison with its own neutral point. 50 is not half a relationship and not a pass mark; it is where a couple lands who answered in the middle of everything. Above it means your two sets of answers converge more than that couple's did. Below it means more of them pull apart than together. The distance from the neutral point carries the information — the digit on its own does not.

Then there are the words printed next to the number. Each band has a caption, running from “a lot is pulling in different directions” through “something to lean on, and something to repair” and “a working couple with clear bottlenecks” to “a strong pairing” and “aligned on almost every axis we measured”. One thing is worth knowing about them: through the working part of the scale the bands are narrow. Neighboring captions are separated by a handful of points rather than by a chasm, and a couple can move from one wording to the next on a change far smaller than the difference between the two phrases sounds. Read the caption as a tag attached to a number, not as a grade, and certainly not as a target to train toward.

Where your compatibility score comes from

Each partner answers for themselves; each of the nine dimensions turns into a personal score from 0 to 100; the two personal scores are combined into a pair score; and eight of the pair scores are added up with fixed weights. The weights are set in advance and are identical for every couple. Trust and Communication carry the most — trust is the precondition for everything else, and communication is the only dimension that describes a process rather than a property of a person, the process a couple uses to work through disagreement. Emotional Maturity, Empathy, Responsibility and Peace of Mind follow. Lightest of the eight are Togetherness & Space and Future Vision, and that is not a slight: they are scored on matching, and a couple who agree there already reach the top without needing much weight to get there.

The ninth dimension, Partner Responsiveness, stays out of the Love Score entirely. It is the one scale where you answer about the other person rather than about yourself, and including it would count one partner's view twice inside a number that belongs to the couple. The whole path from a tapped answer to a finished sentence — including where the formula stops and the model starts — is written out in how AI calculates compatibility.

Why a high score is not a guarantee

The Love Score says one thing only: how well your two sets of answers agree with each other and with the model behind the test. It has never been validated against what actually happened to real couples, and it promises nothing about next year. A couple with a high score can come apart over a subject the questionnaire does not contain a single question about — a move, a parent falling ill, a job in another city. The number describes a configuration, not a fate. If you came looking for the opposite of that promise, the case against forecasting is made in full in can a test predict a breakup.

Why a low score is not a verdict

Two features of the formula matter here, and both of them make a low number mean something narrower than it feels.

The first is a ceiling that cannot be bought out. If a couple has a deep shortfall on one of the critical dimensions — Trust, Communication, Peace of Mind, or, for couples who are past the first months, Future Vision — the final score is capped from above, and excellent results elsewhere do not lift it. That is deliberate: a couple with broken trust and superb domestic logistics should not read “everything looks fine”. The practical consequence is the useful one. A low Love Score frequently points not at “all of it is bad” but at one specific support that has dropped, and the profile across the dimensions shows which.

The second is that some of the scores are not about better and worse at all. Two of the nine dimensions are scored on matching rather than on level: Togetherness & Space and Future Vision. There is no correct value on either. A couple who spend nearly all their time together and a couple who each keep their own trips and their own friends are equally aligned. The pair score comes straight out of the distance between the partners: the average level plays no part, closer answers score higher, and where two answers have become mutually exclusive it falls to zero. A low number there names a topic the two of you read differently — it does not name anyone who failed. Why sameness is only good news on some scales and irrelevant on others is the subject of how similar couples need to be. Two further scales, Communication and Responsibility, are hybrids: most of the score comes from level, a smaller part from how closely the two of you describe the same events. All nine are walked through with examples in the nine relationship dimensions.

Your strengths as a couple: not “how much good we have”

The second number on the screen is captioned “your strengths as a couple”, and it is almost always read as a percentage. It is not one — the % sign is forbidden for this metric in the configuration itself. The metric checks ten specific load-bearing supports: trust as the foundation, conversations that get carried through to a decision, self-regulation in conflict, understanding each other, reliability and contribution, calm instead of control, a matching balance of closeness and space, a matching picture of the future, mutual responsiveness, and a self-image that matches how you are seen. It reports how many of them are actually pronounced in your answers.

The key word is both. For eight of the ten, a support is credited only when the quality is present in both partners — one strong person does not create a support for two. The remaining two, the matching ones, work differently: what matters there is not “high in both” but “the answers converged”, and the support is awarded only if the topic is pronounced for at least one of you. Agreeing about something neither of you has ever thought about is not a resource.

Hence the captions, which run from “hardly any supports visible” through “a few solid supports”, “a steady set” and “a strong resource” to “supports across the board”. The size of the number is not the share of good things in your relationship; it says only that some of the load-bearing structures are pronounced and the rest did not qualify. The practical reading is simple: look at which supports were named rather than at how many. That list is what you have to lean on in a difficult month.

Friction points are not the flip side of your strengths

The third number is computed separately and from its own nine components: a shortfall of trust, conversations that do not get finished, both partners reacting fast, control and separation anxiety, a picture of the future that does not line up, a mismatch on personal space, contribution spread unevenly, two different pictures of the same reality, and one partner carrying the empathy. This matters more than it sounds: friction is not a hundred minus your strengths. Both numbers are often high at once — that is a couple with real supports and one serious divergence. Both are often low, which reads as calm and even, with not much to lean on either.

A tenth component exists in the configuration and is switched off in the current version; there is no point looking for it in your report. The blind spot rests on a measure we have not calibrated yet, and scoring people on an uncalibrated scale is precisely what we criticize other tests for. The remaining nine are renormalized so that the total still adds up the same way.

The bands here are deliberately narrow at the bottom: the distance from “no pronounced risks visible” to “a critical configuration” is shorter than you would expect, because several components firing at once is a system rather than a detail. The top band is not a sentence and not an alarm. It means that working through this alone, one item at a time, is not a realistic plan — which is an argument for a conversation, with each other or with someone outside. And a modest value does not mean “we are on the edge”; it means one or two components fired noticeably and the others did not. The right question for this block is not “how much” but “what”: a friction point is a topic worth discussing before it discusses you.

Why there are two archetypes, and how to read them

Each partner gets an archetype of their own. There are twelve: eight are built on a leading dimension (Open Book for trust, Diplomat for communication, Keystone for responsibility, and so on), three on characteristic pairings of two dimensions, and one, Steady Flame, belongs to a profile with no pronounced leader. All twelve are described in the archetype catalogue.

The thing to know while reading: an archetype describes the shape of a profile, not its height. It is computed from how far your eight traits depart from your own average, which answers the question “what stands out most in this person” rather than “how well is this person doing”. A fixed per-scale correction is applied to those departures before they are compared: by the design of the questionnaire some scales rise more readily than others, and without the correction they would lead more often than they deserve to. The correction comes from a simulation of the model itself rather than from a sample of people, and it does not touch the size of your scores — only which dimension is recognized as leading. So a Diplomat with strong answers and a Diplomat with weak ones are one archetype and two very different people. If nothing stands out — your sides run level, and calling one of them leading would be dressing up ordinary self-report noise as a portrait — the result is Steady Flame, which is neither a rarity nor a prize but an honest statement about an even profile.

The second thing: an archetype does not depend on your partner. It is computed from your answers alone and would have come out the same if you had taken the test on your own. From which follows the point that is most often misunderstood: pairs of archetypes are never “compatible” or “incompatible”, and there is no combination chart anywhere in LoveScore. Two archetypes are a vocabulary for naming the difference between you, not a verdict on it. Compatibility is measured by nine scales, not by two names — the derivation is set out step by step in relationship archetypes.

The alignment index: six scenarios, three of them free

The last block is the alignment index. It covers six concrete joint steps: moving in together, getting married, surviving a renovation, starting a business together, having children, and everyday life together. Each has its own formula — the same dimensions enter different scenarios with different weights. For a renovation, responsibility, communication and emotional maturity carry the most; for marriage, how closely your answers about marriage agree, and trust; for children, maturity, responsibility and agreement about children. The captions are shared across all six, from “low convergence” through “a mixed picture” and “leaning towards each other” to “high convergence”. The comparison worth making is not between one scenario and its band but between your own six scenarios: the spread among them shows which step you have approached with more agreement than the others.

What the index does not measure is the likelihood that you will take the step. It is about how aligned the two of you are as you approach it today. If the topic is live for neither of you, the scenario is honestly marked as not applicable instead of being scored just in case. If your answers on the anchor topic diverge radically and the whole Future Vision scale confirms it, the card gets a ceiling: the number is not allowed to rise into reassuring territory where the divergence is structural. Three cards are shown in the free part, chosen by the same algorithm that does the scoring — no model involved — on informativeness, with one constraint: all three cannot come out below your own average across scenarios. They are picked from five, since everyday life together is deliberately excluded from the free pool and appears there only in the rare case where fewer than three eligible scenarios remain. The rest, and the written explanation of each, are in the paid report.

What compatibility test results cannot tell you

This is the most important part of the reading, and it is worth taking whole. Every line below is a limit of the method rather than a limit of this particular report.

• They do not compare you with other couples. The Love Score is not a percentile. The bands are derived from properties of the model rather than from a sample of real couples, and phrases like “better than most couples” are on the stop-list the generated text is checked against. There is no base for such a comparison.
• They do not predict the future. No score here has been tested against what later happened to anyone. Forecasts and estimates of the odds of separating are blocked by the text validator.
• They do not diagnose. This is not a clinical instrument, and it makes no assessment of anyone's mental health. Clinical vocabulary and attachment-style labels are blocked in the same place.
• They do not rate your partner for you. Item-by-item answers stay private to the person who gave them; the other one sees the couple-level report only.
• They measure a description, not a relationship. This is self-report: the mood of the day, the wish to look better and different readings of the same wording all move the answers. That is why the questionnaire carries attention checks and watches pace and sameness of responses. Under moderate doubt the individual scores are shown as ranges and the text becomes more careful; where the answers are clearly unreliable no report is built at all, because there is no honest report to be made out of random taps.

Which also answers the question people ask most often: will it be the same if we take it again? With the same answers, yes, exactly the same — there is no randomness in the calculation. With different answers, a different number, and that is a property of self-report rather than a fault. It is also why the test is worth taking separately and without comparing notes: an answer given with your partner watching corrupts the data more quietly than a careless one, and no check will catch it.

What to do with the report the same evening

The reading order that works best is the reverse of the obvious one: start not with the Love Score but with the strengths block. Read the list of what is working out loud. It is the one part of the report couples almost always scroll past, and the one part they can act on immediately. Then look at the friction points and pick exactly one — the one with the highest number, or the one whose wording surprised you. One, not three.

Then the conversation, which has one simple rule: discuss what is behind the number rather than the number. “We got a sixty-one” is a dead end. “It says here that the two of us describe our own conversations differently — tell me how it looks from where you are standing” is a beginning. If the figure stung, say so plainly: a reaction to a number is information about a couple too, and often more valuable than the number. Using the score to blame your partner is doubly pointless — it was computed from both sets of answers, and by construction there is no personal fault in it.

And last. How the calculation works — which dimensions enter the score, by what rules two people's answers are combined, and where the boundary runs between the algorithm and the model — is written out on the methodology page. If you do not have a report yet and would like one, you can start here: 47 questions, about seven minutes, each of you on your own device. The basic part — Love Score, strengths, friction points, both archetypes and three scenarios — is free; the full report, with every metric explained and all six scenarios, is a one-time $2.99 for the couple.

Frequently asked questions

Keep reading

LoveScore is an entertainment and educational service for adults 18+. Articles and reports are not a psychological consultation or a diagnosis, and they do not replace working with a specialist.