Skip to content

Right now no language model is switched on: every reading on this site is written from our templates.

How we use AI

Which part of a reading is arithmetic, which part comes from templates we prepared, when a language model is involved, and why we put this on its own page instead of in the footer.

Last updated 2026-09-27

Two layers, and they are different things

Everything on this site comes out of two steps that are easy to confuse, so we separate them here.

The first step is calculation. The second is writing the words about it. The first is always done by fixed rules. The second is done by software from templates we prepared, and involves AI at the moment you ask only where a language model is switched on.

The chart is calculated, not generated

Your birth date, and your birth hour if you gave one, are converted into four pillars using the classical rules: solar terms to find the month boundary, the sexagenary cycle for stems and branches, and a weighting of the five elements that includes the hidden stems inside each branch.

This is arithmetic. It is deterministic, it does not involve a language model, and it could be done by hand with the right tables. Run the same details in another BaZi calculator and the pillars should agree with ours. If they do not, one of the two implementations is wrong and we would like to hear about it.

Where your birth hour is missing we calculate three pillars rather than four, and we label the reading accordingly instead of inventing the missing pillar.

The words come from our templates

The words about your chart come from templates we prepared, which set out what the tradition says about each configuration. Software picks the templates that match your chart and fills in your figures, so the same details always give the same words.

We drafted these templates with the help of AI writing tools.

Our templates are checked automatically against a list of phrases we refuse to publish — promises of outcomes, claims of certainty about the future, and copy that tells a reader their finances are in trouble.

When a language model is switched on

We can switch on a large language model to rewrite parts of a reading within limits we set, working from the same two inputs: the figures the calculation produced and our templates. Anything it writes is checked against the same list before it is shown.

The model is not consulting anything mystical. It produces plausible language about a chart, informed by the tradition we encoded, and it can be fluent and wrong at the same time.

There is no human practitioner

Nobody reads your chart. There is no master, no team of readers, and no photograph of a practitioner on this site, because there is no practitioner to photograph.

This is the point where most sites in this category are quietly dishonest: a named master, a warm biography, and software behind the curtain. We would rather lose the sale than run that.

What this means for you

Treat the reading as a well-informed description of a tradition, not as a message from anywhere. It is offered for entertainment and self-reflection, and it is not financial, medical or legal advice.

Different schools of BaZi read the same chart differently. Ours is one reading among several defensible ones, produced quickly and cheaply, which is exactly why it is free.

If the reading says something that does not fit you, that is useful information about the reading, and you are allowed to put it down.

Why we disclose this

Partly because it is right. Mostly because the alternative is worse: a reader who discovers on their own that the master was software has been deceived, and no amount of good writing recovers from that.

Disclosure costs us something. Research on AI disclosure is consistent that saying so reduces trust slightly compared with saying nothing. Being found out reduces it far more, and permanently.

If you want to ask a human about any of this, write to hello@orientalpower.co. A person answers that address.