Kundli Matching with ChatGPT: Why Gun Milan Needs a Calculator

Families increasingly try the quick route: paste two sets of birth details into ChatGPT and ask for the Gun Milan. The answer arrives in seconds — a score out of 36, a koota-by-koota table, a line about Nadi dosha and a verdict. The short answer to whether you can rely on it is no, not for the score itself. Ashtakoot matching is arithmetic on two exact Moon positions and a set of fixed tables, and a general chatbot asked directly does not do that arithmetic. It writes what a match usually looks like.

That does not make the chatbot useless here. It makes it the wrong tool for one step. This post shows which step, why it matters for a decision as serious as a marriage, and how to check any matching result yourself. If Gun Milan is new to you, the full guide to Ashtakoot Guna Milan explains each koota in depth.

What Gun Milan Actually Calculates

Ashtakoot Guna Milan compares two charts across eight kootas, weighted from one point to eight:

Every one of these comes from a single point in each chart: the Moon's nakshatra and rashi. Get one Moon wrong by a nakshatra and as many as all eight kootas can change at once. That is why the input matters more than anything a reading says about it.

Where a Chatbot's Match Goes Wrong

1. The Moon is estimated, not calculated

The Moon moves about 13 degrees a day and spends roughly one day in each nakshatra. Without an ephemeris, a chatbot places each Moon from patterns in text — and on any birth date where the Moon changed nakshatra, even a careful estimate has two possible answers. A calculator either knows which one applies or tells you the result is provisional. A chatbot typically picks one and presents it as certain, which is the single most common way matching results mislead people.

2. The tables are recalled, and the tables differ

Published koota tables are not identical. The gradations of the Yoni grid, the direction the Gana table runs and the Vashya scores differ between references — enough to move most totals by a point or two. A careful calculator names the reference it follows. Our kundli matching tool follows the open-source PyJHora tables for those three, Muhurta Chintamani for Varna, and the classical rule for Tara and Bhakoot where PyJHora departs from it; for a Sagittarius or Capricorn Moon it also sets the Vashya class by the Moon's degree rather than PyJHora's pada rule. A chatbot recalling tables from memory can blend several references inside one answer, and give a different blend next time.

3. Doshas are reported without their cancellations

Same Nadi scores zero out of eight, which alone can pull a strong match below 18, and couples have been separated on that number. But the texts that define the dosha define its pariharas in the same breath. Following Muhurta Chintamani, Nadi dosha is cancelled when both Moons share one rashi in different nakshatras, or one nakshatra in different rashis or padas. Bhakoot dosha is cancelled when the two Moon signs share a lord or their lords are friends, and Gana dosha when the Moons share a rashi or their sign lords are friends. A chatbot may mention cancellations in general; it cannot tell you which ones apply to your pair unless it has the two calculated Moons in front of it.

4. Mangal Dosha is read from one chart, one way

Mangal Dosha is not a property of a match; it is read in both charts, each from three references — the lagna, the Moon and Venus — and graded by how many of the three Mars afflicts. Mars joined by Jupiter or the Moon cancels it, and when both charts carry the dosha it cancels out (dosha-samya). A one-line "the girl is Manglik" answer has skipped most of this. The Mangal Dosha guide covers the grading in full.

Side by Side

The matching calculation is deterministic: the same two sets of birth details always give the same score, with no AI model in the loop and nothing estimated.

How to Check Any Matching Result

Whatever tool you used, these five checks take a few minutes:

1. Compare both Moons' nakshatras with a calculated chart for each person — a free kundli shows the Janma Nakshatra. If either is different, ignore the score.
2. Look at the koota table, not just the total. A 22 that loses all eight Nadi points is a different match from a 22 that loses points evenly.
3. Ask which cancellations apply, and from which text. "Nadi dosha present" without a cancellation check is half an answer.
4. Check Mangal Dosha in both charts, from all three references.
5. Ask twice. A calculation gives the same 36 points every time. If the score moves, nothing was calculated.

What the Score Cannot Tell You

Even a perfectly calculated Gun Milan describes a traditional calculation, not a prediction about a relationship. It was designed as one input among many. A birth time adds the parts a Moon-only method cannot see: the lagna reckoning of Mangal Dosha, which can change its grade, and the Navamsa comparison — each chart's seventh lord and Navamsa lagna, Venus and Jupiter. Beyond the charts there is everything about two people that no koota measures.

So the sensible order is:

1. Calculate the match with a tool that shows its working — our kundli matching is free and needs no sign-up.
2. Read the doshas with their cancellations before reacting to the number.
3. Interpret with an AI astrologer that has both calculated charts in front of it. From the matching result, the AI Jyotishi reads both kundlis together, with all eight koota scores already computed, and you can question it directly — which kootas failed and whether it matters, what the seventh houses say, and how the timing looks. Your first question is free; see how readings work.

This is the same principle as making a kundli: let software calculate and let AI explain. If you want to see why that order matters for a single chart, read Can ChatGPT make your kundli?. For the marriage question itself — timing, the seventh house and the Navamsa — the marriage prediction page goes further.