Using ChatGPT for your natal chart: what it's actually worth
Pasting your birth date into ChatGPT and asking for your chart has become a reflex. The answer arrives in seconds, it reads well, and it looks right. The problem is not the writing — that part is genuinely excellent — but where the numbers underneath it come from.
What a language model is genuinely good at
Explaining what the eighth house covers, rephrasing an aspect in plain language, comparing two symbolic traditions, answering a vocabulary question: on everything cultural, a large language model is remarkable. It has read more astrological text than any practitioner ever will, and it teaches that material better than most books. If your goal is to understand a concept, it is an excellent tutor.
What it fabricates without telling you
A planetary position is not knowledge, it is a computation. It needs ephemerides, the exact coordinates of the birthplace, the time zone actually in force that day — daylight saving included — then spherical trigonometry for the Ascendant and the houses. A language model does none of that: it emits the most likely-looking degree given what it has read. The result is credible, sometimes close, often several degrees off, and it changes between sessions for the same birth. Since nothing flags that uncertainty, a wrong chart reads exactly like a correct one.
The combination that does work
There is no need to choose between calculation and AI — each is good in its own place. Spikka computes the sky first with a deterministic engine: positions, angles, houses, aspects, fixed stars, progressions, directions, solar returns. Only then does a language model write, taking that chart as its input. The AI no longer has to guess anything; it explains a sky that is already established. You keep the writing quality of a modern model without the risk of an invented chart.
ChatGPT alone and Spikka, line by line
| What we look at | Spikka | ChatGPT alone |
|---|---|---|
| Planetary positions | Computed from ephemerides by a deterministic engine. | Recalled by the model: plausible, never verified against a calculation. |
| Place and time zone | Coordinates and historical time zone resolved automatically as you type the place. | Depends on what the model believes about the place and the daylight saving rule that year. |
| Ascendant and houses | Placidus houses computed to the minute from the birth time. | Often approximated, sometimes skipped, rarely recomputed from the exact time. |
| Reproducibility | Same birth, same chart, every single run. | The same prompt twice can return two different charts. |
| Dated forecasts | Transits, progressions and directions computed against real dates. | Dates are stated, with no underlying computation. |
| Role of the AI | Write and explain a chart that is already computed. | Produce both the data and the prose. |
Comparison made in August 2026, for ordinary use of a general-purpose assistant with no astronomical calculation tool attached. A model that can run code or call an ephemeris library gets far better positions: the gap described here concerns the plain conversation most people actually have.
Frequently asked questions
Can ChatGPT calculate my natal chart?
Not in plain conversation: it produces plausible rather than computed positions. With a calculation tool explicitly attached — code execution, an ephemeris library — the answer becomes exact, but that is not what happens when you simply ask.
Why do its answers change from one time to the next?
Because a language model generates probable text, not a unique result. An astronomical calculation has exactly one answer for a given birth. It is the easiest thing to test yourself: ask twice and compare the degrees you get.
What if I give it my exact positions myself?
Then it becomes excellent. Hand it a correct chart and it will interpret it with real nuance. That is precisely how our assistant works: it never receives your raw birth data, only the chart already computed from it.
Spikka uses AI too — what is the difference?
The division of labour. Here the AI has no influence on the data: it picks no degree, no house, no aspect. It puts into words a chart produced by a calculation engine, and says when something is missing instead of filling the gap.
Can it give me forecasts for the year?
It will give you dates, and they will usually be wrong, because no transit, progression or direction was computed to produce them. A forecast means dating the real passage of a planet over a real point of your chart — two values that must be calculated, not estimated.
Do I have to pay to see my chart on Spikka?
No. Creating your space and computing your chart are free, with no card required. Written reports and the chat subscription are options, detailed on the Pricing page.