How to choose an AI BaZi product: eight questions you can check
Judge an AI fortune-telling product by eight checkable items rather than its marketing: how it charts, how it shows evidence, whether it addresses the past, how transparent its conventions are, how it treats the birth hour, what it refuses to do, how it handles privacy, and how it prices.
Direct answer
Choose an AI BaZi product by asking whether an engine casts the chart, whether conclusions cite specific stems and branches, whether its description of the past can be checked, and whether its conventions and limits are stated openly. Price and polish matter only after those questions are answered.
First ask whether it lets you check anything
AI fortune-telling products increasingly look alike: enter birth information, wait a few seconds, receive a long reading. What actually separates them is not the prose style but how much of the work behind that prose can be inspected. Before the words 'master', 'secret lineage' or 'precise' carry you away, look at the process.
There are three layers to examine. The first is charting: are the four pillars calculated by a deterministic engine, or worked out by a language model during the conversation? The second is interpretation: can each conclusion land on specific stems and relationships, and can it be questioned? The third is boundaries: does the product state clearly what it does not do, and does it admit uncertainty when it exists? Only after all three pass does price and experience deserve attention.
Two more dimensions are often overlooked: privacy and pricing structure. Birth information is sensitive data, and the terms describing how it is stored and used are worth reading before payment. Price can reflect coverage and service shape, but it is not a direct measure of credibility. What a free tier lets you see of the chart and its evidence often says more about the product's logic than its marketing does.
This article does not score any particular product and does not imply that expensive is better. It offers a set of questions. Take these eight to a trial, and the signals of pass and fail will be far more stable than a feeling of whether the reading seemed accurate.
Comparison table: eight evaluation dimensions
| Dimension | Question to ask | Warning sign |
|---|---|---|
| Charting mechanism | Are the four pillars calculated by an engine or worked out by the model? | The result differs each time, and the charting process cannot be shown |
| Presentation of evidence | Which stem or branch does each conclusion land on? | Verdicts only; the explanation changes when you ask why |
| Past check | Does it describe earlier stages for you to compare? | It talks only about the future and skips the testable part |
| Transparent conventions | How are true solar time, the zi hour and solar terms handled? | Conventions are never mentioned, as if only one method existed |
| Birth-hour handling | Are boundary cases flagged? | They are handled silently, so the user never learns the chart may change |
| Stated limits | Does it explicitly avoid health, disaster, investment and legal topics? | It promises everything and uses fear to drive payment |
| Privacy and data | How is birth information stored and used? | Vague terms and requests for unrelated permissions |
| Pricing structure | What separates the free and paid tiers? | It promises to change your fate after payment and uses anxiety to drive upgrades |
The first six dimensions decide whether the product can be trusted at all; the last two decide whether using it feels safe. A clear failure on any one is enough reason to choose another tool.
How to verify: test in this order during a trial
Step one: ask for a complete chart first.
The four pillars, hidden stems and Ten Gods should be displayed directly, not buried inside a long article. A product that cannot produce a complete chart, or produces a different one each time, is not worth further discussion. The chart is the shared starting point, not decoration.
Step two: test boundary dates deliberately.
Try a birthday on the day of Start of Spring, a day when a sectional term changes, and a birth after 23:00. Boundaries expose on-the-fly calculation most clearly: an engine gives a stable result and explains its convention, while a model that calculates in the conversation tends to drift at exactly these points. If your own birthday is near a boundary, this step matters even more.
Step three: ask one conclusion for its evidence.
Pick any judgement in the reading and ask which character or relationship it came from. A claim that lands on specific stems and branches can be checked. An answer that circles back to 'the overall feeling' or produces a new paragraph instead is a reason to discount the reading.
Step four: check its description of the past.
Do not add details about your recent situation. Read what it says about earlier stages, relationship patterns and turning points, then compare that with your own experience. This site has no separate feature called a 'past event check', but any product can be tested this way by the user; a reading that misses widely on the past should not be used directly to discuss the future.
Step five: inspect the explanation of conventions.
See whether it states whether true solar time is on, which convention governs the zi hour, and how solar-term boundaries are handled. The convention does not have to match another product, but it must be stated. Without that, you cannot tell whose choice produced a difference.
Step six: read the limits and privacy terms.
Check whether it explicitly avoids medical diagnosis, disaster claims, investment and legal conclusions, and whether it explains how birth information is stored and used. The clearer the limits, the less likely you are to be pushed around when you are emotionally vulnerable.
Step seven: compare the free and paid tiers.
A free tier should at least show the chart and part of the evidence. Paid tiers should differ in coverage and follow-up depth, not in whether you are allowed to believe the result. If the core chart is hidden in the free tier, the paid reading also lacks a reference point for credibility.
Step eight: check the cost of leaving.
How easy it is to cancel a subscription, delete birth information and export records is part of product discipline. A product that puts obstacles in the way of leaving is unlikely to be restrained about the limits of its readings.
How we approach it
Applying these eight questions to this site, we give the checkable location for each and also say where we are not perfect:
Charting is done by an engine. The four pillars, hidden stems, Ten Gods and Major Cycles are calculated by a deterministic engine, and recasting the same birth information returns the same result. The AI only reads the finished chart; it does not take part in charting or modify it.
Conclusions must land on stems and branches. Judgements in a reading should correspond to specific characters and relationships. When strength is near a threshold, the boundary is marked; when support-and-restraint and climate-adjustment layers differ, the divergence is stated. You can question each item rather than accept or reject a whole paragraph.
The description of the past is there for you to check. The structural description covers earlier stages and relationship themes so you can compare them with your own history. We do not have a separate feature called a 'past event check', and we do not tell users that a particular experience must have happened. The checking is yours to do.
Conventions and limits are written in the open. True solar time can be switched on or off, the zi hour changes the day at midnight, the year pillar turns at Start of Spring and the month pillar at sectional terms; boundary cases are explained in notices. Medical diagnosis, disaster claims, investment and legal conclusions are excluded across the board.
Payment does not promise to change your fate. Paid tiers differ in coverage and follow-up depth. We do not promise to change your destiny and do not use fear to drive upgrades. Birth information is handled under the privacy policy; whether a paid tier is worth it should be decided first by the quality of the chart and evidence.
We admit the limits of our abilities. This is an AI reading. It does not replace human consultation, counselling, medical care or other professional advice. There is no same-chart search, side-by-side multi-chart view or automatic hour detection. Stating these limits is more useful to your judgement than pretending to be omniscient.
Limits and disclaimer
This article offers a method for evaluating products. It does not recommend or disparage any particular product, and it does not promise that following it will produce an 'accurate' reading. Fortune-telling products can only offer interpretation within a traditional framework. They do not replace medical, psychological, legal, investment or other professional advice. For major decisions, rely on real-world conditions and qualified advice.
Frequently asked questions
What is the difference between free and paid products?
Usually coverage, follow-up depth and service shape, not credibility. Whether the chart is complete and the evidence visible should be judgeable in the free tier. If the free tier gives only an unsourced verdict, paying more does not necessarily make the product more checkable.
Is a longer reading always better?
No. Length is not information. A good reading lands on specific stems and relationships; a long text that repeats conclusions and piles up adjectives is harder to check. Look at the evidence before the word count.
Is the phrase 'AI master' credible?
The word 'master' is a piece of identity packaging, not a testable indicator. Judge a product by its charting mechanism, presentation of evidence and stated limits, not by what it calls itself. The more it emphasises status, the more you should ask where the evidence is.
How can I judge data privacy?
Look at whether the terms state how long birth information is stored, what it is used for and how it can be deleted, and whether the product asks for permissions unrelated to its function. Be cautious about entering detailed birth information into a product with vague terms and a difficult exit.
What should I test during a trial?
Test boundary dates first: Start of Spring, a sectional-term change, and a birth after 23:00. Then check whether the four pillars are stable, whether conventions are stated, and whether follow-up answers have evidence. Also cast your own birthday once and compare its description of earlier stages with your experience.
Why do some products avoid showing their charting process?
Possibly because the model calculates the chart in the conversation and results are unstable, or because the product prioritises narrative over calculation. Whatever the reason, if you cannot see the chart, you cannot check the starting point and can only accept the conclusion.
What is the single most useful question to ask before paying?
'Which character did this conclusion come from?' If the product answers that consistently, it has a checkable structure. If it can only say 'the AI considered everything', what you buy may be longer text rather than more reliable analysis.
Where are this site's limits, and what does it not do?
It does not predict specific events, diagnose illness, make disaster claims, or give investment or legal conclusions. It does not offer same-chart search, a side-by-side multi-chart view or automatic hour detection. True solar time, zi-hour and solar-term conventions are stated, but they are not the only standard.
Related comparisons
Updated 2026-10-20