I have, on purpose, dabbled with Chat GPT and Claude for several birth chart analyses. This was, of course, done after the thorough exhaustion of my own human interpretation.
Without any variance, the LLM models have presented an over-flattering synthesis of the horoscopes (perhaps, on purpose?) over time. I have now reached a conclusion that the models tend to overplay the good parts and hugely underplay the bad parts. I believe this is perhaps because the models are not designed to cause any kind of distress to the user in order to deflect any public outrage or potential lawsuit.
What I have also formed an opinion about certainly now backed by factual incidents, is that the model is a rolling stone – you never really know which side it is on. The model will start by providing an analysis which you, falling prey to your human nature, will retrofit. The moment you, as the user, provide a contrarian fact, the model will promptly switch gears and retrofit your analysis or shamelessly apologize and make blatant corrections.
Basically, the model is designed to be a user-pleaser, and it does so excellently. People have become self-taught astrologers and have begun to predict their own lives. They can’t be blamed though because access is so damn easy and the replies are incredibly fast.
As a practicing astrologer, I struggle to apply all the concepts that I have learnt over the years while studying a birth chart as they are extremely pedantic to begin with, leave alone their application. However, both Chat GPT as well as Claude, meritoriously apply all the rules and guidelines available in any decent astrology book, with fine precision. Contextually speaking, there are 9 planets, 12 houses, 12 zodiac signs and 27 nakshatras. So, there is a practical limitation to how much the human mind can process for any one such combination in a birth chart. But the LLM’s were designed to that anyway, right? Where they start failing is that almost invariably, they will forget to consider “yogkarakas” or “badhakesh” or “exchange yogas” and many other important considerations. This causes extreme frustration in case I am looking to sharpen my analysis because the entire interpretation of the birth chart changes if any one of these become an oversight for the LLM.
Nevertheless, I personally believe that AI should strictly be used only by a professional astrologer to eliminate outlandishly flattering analyses. It should always be used as a tool even by the experts and never as a replacement of their original thought process and instinctive reading abilities. For a beginner, the usage is a strict “NO”. Only when one can successfully determine a right interpretation from a wrong perspective, should one dabble with the LLMs.
