This is the part most AI astrology apps quietly skip, so it is worth being specific about what is calculated and what is interpreted.
A language model asked to “cast a birth chart” will produce something that reads like a chart. The planets will be in plausible signs, the degrees will look precise, and a good proportion of it will be wrong — because the model is predicting text, not computing positions.
CosmicFlow never asks it to. Every planetary position, house cusp, lagna, nakshatra, dasha period and Guna Milan score is calculated by the Swiss Ephemeris — the same astronomical engine professional astrology software has relied on for decades, derived from NASA’s JPL ephemeris — using the Lahiri ayanamsa for all Vedic work.
The AI receives that computed chart as fact and writes the interpretation. It decides what a Saturn return in your tenth house might mean for your career; it does not get to decide where Saturn is. If our arithmetic disagreed with a professional astrologer’s software, that would be a bug we would fix — not a difference of opinion.
The same applies to matchmaking. Ashtakoot Guna Milan scores are computed from both Moon nakshatras with the classical dosha-cancellation rules applied, not estimated from sun signs. It is why a match can score differently here than on sites that only compare star signs.
Eighteen guides read that same computed chart, each through one tradition’s technique — Vedic dashas, Hellenistic time-lords, Chinese BaZi pillars. Ask two of them the same question and you should get two different readings. That disagreement is real, and it is the point.
Meet the 18 guides →