How it works

A clearer way to discover compatible people

Twinbat combines what you tell it with signals from how you use the service. It explains why a person appears relevant, and lets you control which data contributes to that.

What your profile is built from

  • Identity. Your basic details and confirmation that you are 18 or over.
  • Intent. What you are open to — friendship, relationships, professional connections, mentorship, Circles — and any preferences you choose to give.
  • You. Interests, values, goals, life stage and how you prefer to communicate.
  • Birth details. Optional. Date, and if you have them, exact time and place. If the time is unknown, anything that depends on it is marked unavailable or lower confidence — never guessed.
  • Location. Optional and approximate, for local discovery. Entering a city by hand always remains available.
  • Behaviour. Saves, passes, completed comparisons, Circle participation and messages that were actually exchanged both ways.

How matching actually runs

Hard eligibility rules run first: age, blocked accounts, excluded intents, location settings and privacy controls. Only what survives that is ranked, using the discovery mode you chose.

A paid plan can widen your filters or deepen an insight. It never bypasses a safety, privacy, consent or eligibility rule. There is no amount of money that shows you somebody who blocked you.

Every result shows why that person appeared, which score components were available, and whether missing or stale data reduced confidence. You can reset what Twinbat has learned about your preferences without deleting your account.

Where the astrology stops and the AI starts

The astrology engine calculates placements, relationships and scoring inputs from the birth data you provide, using versioned rules. That calculation is deterministic: the same inputs and the same version produce the same chart, every time.

The language model receives only approved, structured results and turns them into readable explanations. It does not calculate a chart from memory, infer a missing birth time, invent placements, or claim certainty about the future. That separation is the whole design: a model that can write a plausible chart is a model that will, and you would have no way to tell.