How we check what we tell you

To stand behind an interpretation, two things have to hold: the calculation must be right, and the interpretation must stay faithful to it. This page explains how we test both. We also write up the attempts that failed, because showing what we don't know is part of earning trust.

Calculation: same input, same result, every time

Planetary positions, houses and aspects are calculated with the Swiss Ephemeris library, which we use under its commercial license. The library's files are used unmodified, and every test run automatically checks that they haven't changed.

When the calculation engine was built, every planetary degree, the rising degree and the Midheaven (MC) of a sample chart were compared one by one with the results of a widely used professional astrology program. That chart's values are pinned in the automated tests: if a change shifted a planet by even a millionth of a degree, the tests would catch it.

Some calculations are also redone on every test run with published astronomy formulas that are independent of the Swiss Ephemeris, and the results are compared:

• The Sun's position: across hundreds of dates spread between 1900 and 2100, the difference must stay under 1 arcminute.
• The rising degree (Ascendant) and the Midheaven (MC): across hundreds of places and times between 60° north and 60° south, the difference is at most 2 arcminutes.
• The rotation of the sky (sidereal time): this calculation, which the astrocartography lines rest on, is also checked against an independent formula.

However accurate the calculation: your birth time

Your rising sign and houses depend on your birth time. The rising degree moves about one degree every four minutes on average. So the result can only be as accurate as the time you enter, and the times on official records are often rounded.

If your time is approximate or unknown, you can say so. The app then restricts the time-dependent sections and tells you clearly that it has done so.

AI: interpretation only, then a review

AI doesn't calculate any degree, aspect, house or date. It is handed data that has already been calculated: each planet's sign and house, and each aspect's type, orb and dates. It is explicitly told not to add anything it wasn't given and not to state certain outcomes.

Every interpretation written about your chart passes through an independent second AI review before you see it. The review looks for six kinds of error, each one an error we actually ran into during development:

• Stating a sign that isn't in the data (for example, working out the rising sign from its degree and getting it wrong).
• Merging observations from different dates into a single sentence.
• Adding a planet, aspect or date that isn't in the data.
• Speaking with a certainty astrology can't offer.
• Contradicting in one place what the text says in another.
• Stock phrases that don't rest on the calculated data and would fit anyone.

The limits of the review

If the review finds a problem, the text is corrected once. Which interpretations needed a correction is recorded, so we can track which errors come up most often.

The review reduces errors; it doesn't eliminate them. If a temporary fault stops the review from running, the interpretation is still shown - rare, but possible. That's why every interpretation comes with a “Why?” section: you can check the planet, aspect, orb and date behind an interpretation yourself.

Birth time: we tried, and couldn't find it

Working out a birth time from the dates of past life events is called rectification in astrology, and it's a common practice. Before putting it in the product, we tested it blind on people whose birth times are known from records. The method made its guesses without seeing the real time; the real times were opened afterwards. The test plan and the success criteria were written down and recorded before any results were seen.

In none of the tests we ran in October 2026 did the method find the birth time better than chance:

• First attempt: in a test group of 19 people, the guess landed within 30 minutes of the real time for 1 person. A random pick would be expected to do this for about 0.8.
• Extended attempt: in a test group of 170 people, ranking candidate times by life events didn't place the real time any higher than a random ranking.
• 66 people whose birth minute is documented: the gap between the top-scoring minute and the real minute was about 7 hours at the median. That's the same level as guessing at random.
• Two separate methods based on house rulers (3,600 setting combinations tried in total, 194-person test): the correct rising sign came first 8.8% and 6.7% of the time. A random pick would be expected to do this 8.3% of the time.
• It's a common belief that the rising sign shows in a person's appearance. We also tried finding the rising sign from portrait images (1,198-person test): no usable signal was found.

What do these results mean?

This doesn't mean “rectification can never work.” The people tested were mostly well-known figures and the events came from family and career records, so the results don't carry over directly to everyone. But we found no result that would support saying “we'll find your birth time.”

That's why Birth Time Discovery isn't offered as a birth time prediction. For each candidate time, it shows in ranked order which aspects form on the dates of the events you've lived through; you, or your astrologer, decide which one to use. The tool's screen also tells you how close a time picked at random from your window would be likely to land.

Past echo: a reminder, not a prediction

Period cards sometimes show a “Past echo” note: if the sky pattern of an upcoming period was also present on the date of a life event you entered earlier, we remind you of it. This note is produced by a fixed rule, not by AI.

The birth time tests above showed how hard it is to draw a reliable link between event dates and the sky. So an echo is not a prediction, only a reminder: the same pattern being present then doesn't mean the same thing will happen again.

Your answers to “How did this period go?” on period cards are reviewed in aggregate to understand which notifications are useful to you. A period and an event falling at the same time doesn't show that one caused the other, and we don't present these answers as evidence.

Known limits

• Astrology is not a scientifically validated method of prediction. A correct calculation doesn't mean an interpretation will come true.
• Astrologers disagree among themselves on things like house systems and orbs. The app makes a default choice. You can pick the house system when you create a chart and change the orb sensitivity later.
• If you find something wrong or missing on this page, write to us via the Contact page. We'll fix it and update the page's revision date.