What language models are actually doing
A language model produces text that resembles a correct answer. It is not performing astronomical calculation, and it is not looking anything up unless it has been connected to a tool that does.
That distinction is invisible in the output, which is the whole problem. A chart generated this way arrives formatted, confident and complete, with degrees to two decimal places. It looks exactly like the output of a calculation because it is text shaped like the output of a calculation.
Some of it will be right. The parts that are wrong will look identical to the parts that are right.
The four failures that actually matter
1. Ayanamsa, the biggest one
Vedic astrology measures planetary positions against the fixed stars. Western astrology measures them against the equinox. The gap between the two is called the ayanamsa, and it is currently around 24 degrees.
Twenty-four degrees is most of a sign. Apply the wrong system and roughly four people in five will get the wrong rashi for most of their planets.
This is the failure that does the most damage, because the resulting chart is internally consistent, formatted correctly, and completely different from your actual kundli. If you have ever been given an AI reading that felt like it was describing someone else, this is the most likely explanation. What is a rashi covers the sidereal and tropical difference properly.
2. Time zone and daylight saving
Historical time zones are genuinely hard. India moved between offsets before standardising. Britain has had British Summer Time, and during the Second World War had Double Summer Time. Many countries have shifted their rules repeatedly.
A model asked for a chart from 1974 in a particular city has to know what the local offset was on that date, not what it is now. It often does not, and it does not say so.
An hour of error is enough to change the rising sign, which changes every house in the chart.
3. The rising sign
The lagna or ascendant advances through all twelve signs in roughly twenty-four hours, which means it changes about every two hours. It sets where house one begins, and therefore which house every planet occupies.
Because the rising sign depends on both the exact time and the exact latitude and longitude, it is the single most error-prone value in the whole chart, and it is the one everything else is built on. A chart with the right planets in the wrong houses will produce a fluent reading of a life that is not yours.
4. Retrograde status
Whether a planet was retrograde on a given date is a straightforward ephemeris lookup and a common invention. It matters because retrograde condition changes interpretation, and because getting it wrong is a clean tell that no calculation happened.
Five checks, about two minutes
You do not need to be an astrologer to catch most of this.
One. Ask which ayanamsa was used. The answer should name one, most commonly Lahiri. "Sidereal" alone is not an answer, and a vague reply means no correction was applied.
Two. Check the rising sign against a calculating tool. Generate the same chart on any proper kundli calculator and compare the ascendant. If it differs, stop; nothing downstream is reliable.
Three. Check the Moon. The Moon moves about thirteen degrees a day, so it is the most sensitive to a date or time error and the easiest to verify. If the Moon's sign disagrees with a real calculator, the chart is wrong.
Four. Check one retrograde. Pick Mercury or Saturn and verify the retrograde status for your birth date. It is a binary fact and it is quick to confirm.
Five. Ask it the same question twice, in separate conversations. Genuine calculation returns the same numbers every time. If the degrees drift between attempts, they were generated rather than computed.
Our own kundli tool calculates rather than generates, which makes it a reasonable thing to check against.
Where AI is genuinely good at this
This is not an argument that the technology is useless, and pretending otherwise would be as unhelpful as overclaiming for it.
It is very good at explaining concepts: what a yuti is, what the 7th house traditionally covers, how a dasha system works, what the difference is between Vedic and Western astrology. For learning vocabulary it is faster and more patient than most books.
It is good at summarising traditions and at telling you what different schools say about the same placement.
It is unreliable at calculation, and therefore at anything that depends on the exact chart: which house a planet is in, what is aspecting what, when a period runs. The correct division of labour is to let software calculate and use the model to explain.
The deeper limit, beyond arithmetic
Suppose the chart is correct. There is still something a model cannot do, and it is worth naming.
A reading is not only interpretation. It is a judgement about what to say and what not to say, made by someone accountable for the effect. Which findings matter to this person. Which are true and would only frighten them. When the honest answer is "the chart does not support a clear view on this". When the answer is "see a doctor", and when to refuse a question entirely.
A model asked about lifespan will often answer, because it is generating a plausible response to a question. A practitioner refuses, because the refusal is part of the practice. Reading health in a chart, and its limits sets out where those boundaries sit and why they do not move.
That is not a claim that people are mystically better readers. It is a claim about accountability, which the model does not have.
Before you conclude your chart is bad
If an AI reading described a life you did not recognise, the most likely explanations in order are: the wrong ayanamsa, a time zone error, a rising sign error, and only then anything about you.
The same applies to human readings. If a chart says a period was strongly activated in a year when nothing at all happened, suspect the recorded birth time before concluding anything unusual about your life. Why birth time accuracy matters explains what an error costs and why a stated unknown is far better than a plausible guess.
A BNN birth chart reading starts by checking the chart against events that have already happened, which is the one test neither a model nor a confident astrologer can talk their way around. What actually happens in a consultation describes how that works.