Last updated: 10 August 2026
You are interacting with an artificial intelligence system
The tarot readings, horoscopes, natal charts and compatibility analyses you read on Mago Lucas are written by an artificial intelligence.
There is no card reader on the other side. There is no astrologer proofreading the text before you see it. No person comes between your question and the answer that appears on your screen.
This page explains how it works, what is generated by the machine and what is not, and what the limits of what you read are. We publish it because the European AI Act requires it (Art. 50), but above all because knowing how something works lets you use it better.
What is AI-generated and what is not
This is an important distinction and we want to get it right, because the two are easy to confuse.
Generated by artificial intelligence
- The text of the tarot reading: the interpretation of the cards, how they connect to your question, the conclusions.
- The text of the horoscopes, both daily and personal.
- The text of the natal chart and the astrological portrait.
- The text of the couple compatibility analysis.
- The classification of your question: a model reads the question and assigns it a category (love, work, money…) and a type, in order to choose which card spread to use.
All of that is language produced by a statistical model. The model knows nothing about you or your future: it builds a plausible text out of the information we give it.
NOT generated by artificial intelligence
- The cards drawn. Cards are dealt by our server's random number generator, the way a deck is shuffled. The model does not choose the cards: it receives them already drawn and interprets them. If you get The Tower, it is not because the AI decided The Tower suited your question.
- The positions of the planets. They are calculated with the Swiss Ephemeris, the same astronomical library used by professional astronomy and astrology software. It is mathematics: given a moment and a place, the position of Mars is what it is. Nothing about it is probabilistic and no language model is involved in the calculation.
- The sign, the rising sign and the astrological houses. These follow arithmetically from the positions calculated above.
- The card images and the traditional meanings attached to each card: they are editorial content written by people and stored on the site.
In short: the starting data is calculated, the interpretation is generated. The astronomy behind a natal chart is exact. What is made of it is not.
Which model we use
We use a large language model (LLM) supplied by a third-party company, through its paid programming interface. The provider is currently OpenAI and we use models from the GPT family.
We deliberately do not name a specific version here, for an honest reason: these models are updated and replaced often, and a version written on this page would be wrong within weeks. If we changed provider or model family, we would update this page.
We do not train models of our own and we do not modify the provider's model. We give it written instructions (prompts) describing the tone and structure of the text, along with the cards drawn or the planetary positions.
How your questions are used
When you ask the tarot a question, the text you write is:
- sent to the model, together with the cards drawn and the language you chose, to generate the reading;
- stored in our database along with the reading;
- published, because tarot readings are public by default.
We do not send the model your email, your surname, your IP address or your payment details. The model provider cannot connect the question to you.
Under the contractual terms of the paid API we use, the text sent is not used to train the provider's models. The details, along with the question of data transfers to the United States, are in our Privacy notice.
A practical piece of advice: write your question without health details, without other people's names and without anything that would identify you. There is no need: a short, general question gets exactly the same quality of reading, and what you do not write ends up nowhere.
The limits: what this service cannot do
The text can be wrong
Language models work by predicting, word after word, the most plausible continuation. They consult no knowledge base and verify nothing. In practice:
- They can make things up. A model can state a falsehood with complete confidence. The jargon term is "hallucination".
- They can be inconsistent. The same question asked twice gets different answers, which may even contradict each other.
- They sound more certain than they are. The fluent, assertive tone is a feature of the style, not a measure of reliability. A well-written text is not a true text.
- They can reflect the biases present in the data they were trained on.
There are no verifiable predictions
Mago Lucas does not predict the future and does not claim to. No reading contains verifiable information about what will happen. Cartomancy and astrology are not scientific disciplines: what we offer is a prompt for reflection in narrative form, not a source of information.
If a reading resonates with you, it is because it speaks of common human themes in a way that fits many situations. That is a well-known effect, not a sign that the system knows something about you.
It is entertainment, and nothing else
The content of this site is not medical, psychological, legal, financial or investment advice, and does not replace a qualified professional. Do not use it to make decisions about your health, your money, your legal disputes or your relationships.
If you are going through a hard time, talk to a real person: a doctor, a psychologist, a helpline. A text generated by a computer cannot help you with that, and we will not pretend otherwise.
No automated decisions about you
The system takes no decision concerning you.
It does not score you, does not assess your personality, does not decide whether you are trustworthy, does not sort you into categories affecting what you may or may not do. Nothing it generates produces legal effects concerning you or similarly significantly affects you within the meaning of Art. 22 GDPR.
The only automated classification concerns the question, not the person: it is used to choose the card spread and to recognise sensitive topics.
The safeguards we have put in place
- A filter on sensitive questions. An automated check recognises questions asking for a medical diagnosis or serious legal advice, questions about the private life of identifiable third parties, and questions hinting at harm to oneself or others. In those cases the model is instructed to answer carefully, without making diagnoses or giving legal opinions, and to suggest speaking to a professional; the reading is also flagged for human review.
- No practical advice on health, money and legal matters. The instructions given to the model explicitly forbid it.
- No manipulation. We use no subliminal techniques, we do not exploit vulnerabilities related to age or circumstance, and we do not push purchases by playing on fear. The credit cost of each service is shown up front, and there are no subscriptions or automatic renewals.
- Adults only. The service is for people aged 18 or over.
- Human review on request. Any piece of content can be reviewed by a person: just report it.
These safeguards are automated checks and instructions given to a statistical model: they reduce problem cases, they do not eliminate them. If something slips through, we want to know.
How to report problematic content
If a reading, a horoscope or any other generated text strikes you as offensive, discriminatory, dangerous, seriously wrong or as being about a real person, tell us.
Write to [EMAIL SEGNALAZIONI] with:
- the address of the page (the full URL);
- what is wrong with it, even in one line.
What we do: we read every report, we reply within a few working days and, if the content is problematic, we make it private or remove it. If the problem recurs, we correct the instructions given to the model.
You can report content even if you have no account and even if the content is not yours.
To have a reading you generated removed, or one containing your data, you need give no reason: write to [EMAIL PRIVACY] with the address of the page.
In one line
The cards are dealt at random, the planets are calculated, the text is written by a machine, and it is entertainment. Take it for what it is.
Contact
[RAGIONE SOCIALE], [INDIRIZZO], VAT [P.IVA].
Content reports: [EMAIL SEGNALAZIONI].
Personal data: [EMAIL PRIVACY].
See also: Privacy notice · Terms of use · Cookie policy