Artificial intelligence already writes texts, creates images, composes music, summarises documents, and answers practically any question. And now it also dares to tackle the future. Tarot has not escaped the AI revolution, and there are increasingly more services capable of generating an interpretation of the cards automatically based on the user's questions. Not all sites have taken this path. At tarotamigo.com, for example, they have opted to maintain a traditional system of interpretations crafted by tarotists and forgo artificial intelligence to generate the responses to their readings.
The difference may seem small to someone simply selecting cards on their phone screen, but it completely changes what happens behind the scenes. Tarot Amigo explains that the responses to their readings have not been created using artificial intelligence, but rather from interpretations previously made by their tarotists, taking into account each card and its position. At the other end are the new applications that rely on generative models capable of constructing a different reading in a matter of seconds based on the query entered by the user.
When the tarot reader becomes an algorithm
The operation is straightforward. The user poses a question, chooses one or more cards, and the artificial intelligence receives that data along with the traditionally attributed meaning of each arcana. From there, it constructs a text that relates the cards to the situation presented.
There are already applications that openly advertise AI-powered tarot readings as one of their main attractions. The phenomenon has grown even enough to spark academic interest, with research studying how both enthusiasts and professionals in tarot are incorporating generative models into their interpretations.
AI also has an obvious advantage from a technological standpoint: it can generate millions of different combinations and adapt the language to each query. It does not need to be limited to a pre-written text for The Sun, The Tower, or The Lovers. It can relate one card to another, incorporate the question posed, and deliver a response that appears tailored in seconds.
Precisely there begins the problem.
A personalised response does not mean it understands the questioner
An artificial intelligence does not know what will happen tomorrow. It also does not truly know the person in front of the screen. What a generative model does is produce a response based on the patterns learned during its training and the information it receives at that moment.
However, the result can have an extraordinary appearance of certainty.
If someone asks whether their partner will leave them, if they will get back together with someone they ended a relationship with, or if they will secure a particular job, the system can generate an extensive, coherent explanation that is completely tailored to the problem posed. The better the model, the more convincing it can be.
The issue becomes particularly relevant because many queries related to tarot occur precisely around emotionally sensitive matters: romantic relationships, family problems, job uncertainty, or decisions about the future.
It is not, therefore, that an algorithm “reads worse” or “better” the cards. The problem arises when the user attributes to an automatically generated response a capacity for knowledge, understanding, or certainty that the system does not actually possess.
AI does not see the future, even if it seems to know a lot about us
There is also a paradox. An artificial intelligence can offer a reading that seems even more personalised than an interpretation written previously by a person.
It is enough to give it context.
“My partner left me three months ago, now they are writing to me again and I drew The Judgement, The Lovers, and The Tower. Will we get back together?”
A language model can use each of those elements to immediately construct a narrative perfectly connected to the situation. It can talk about second chances, emotional decisions, breakups, or sudden changes because all those concepts are associated in one way or another with the information provided.
But just because the response fits does not prove that the system knows the future.
Research on the use of artificial intelligence applied to tarot precisely shows this peculiar relationship between interpretation, ambiguity, and automatic generation of meaning. Some users are already turning to these systems to seek alternative interpretations, resolve doubts, or expand their own readings.
The return of something that the internet seemed to have left behind: knowing who is responding
The massive arrival of artificial intelligence is also provoking a discussion that goes far beyond tarot: transparency.
Until a few years ago, it was relatively easy to distinguish between a query made by a person and an automated response. Generative models have blurred that boundary.
A text produced by AI can use colloquial expressions, show apparent empathy, recall previously mentioned elements, and construct an explanation that seems to have been written specifically for a single person.
European regulation of artificial intelligence has also begun to focus on this issue. Transparency obligations seek, among other aspects, to ensure that users know when they are directly interacting with certain artificial intelligence systems.
In areas such as tarot, astrology, or esoteric predictions, that transparency can be particularly relevant. It is not the same to read a previously written interpretation by a tarotist than to receive a text produced at that moment by a generative model. Neither should be confused with a scientifically proven prediction, but the user should know how what they are reading was generated.
Two very different ways to bring an ancient practice online
Technology did not arrive at tarot with generative artificial intelligence. Long before, there were already digital readings where a program selected cards and displayed previously written interpretations.
That model continues to be valid.
Behind it can exist a huge amount of prior work. Each arcana can change its interpretation depending on whether it appears upright or reversed, its position within the reading, the cards that accompany it, or the matter being asked about. Digitising that system means introducing all those possibilities beforehand so that the program can later show the one that corresponds.
Artificial intelligence proposes something different. Instead of retrieving only a previously prepared interpretation, it can construct a new one using all the available elements.
It is more flexible, faster, and offers virtually unlimited possibilities. But it also raises a question that until recently hardly needed to be asked when someone did a card reading online: did a person ever write this interpretation, or did a machine just create it?
Tarot faces the same dilemma as the rest of the internet
In reality, what is happening with the cards is simply a small sample of a much larger change.
Media, search engines, customer service, education, artistic creation, and dozens of sectors are trying to establish where it is useful to introduce artificial intelligence and where it remains important to preserve human intervention.
Tarot now finds itself facing its own version of that dilemma.
There will be those who prefer to receive an interpretation generated specifically for their question and those who place more value on knowing that the meanings presented before them were thought out and written previously by a person.
And both models will likely coexist.
What will probably become increasingly important is knowing how to distinguish them. Because artificial intelligence can interpret the cards, relate them to our question, and construct a surprisingly convincing story.
The only thing it still cannot do is know what will happen tomorrow.
You may also like
