In a year when almost every consumer product has gained a chat assistant, a recommendation engine or a “personalised for you” label, a small corner of the web is moving in the opposite direction. A number of online oracle and card-reading platforms are now stating, in plain language, that no algorithm, no artificial intelligence and no simulated randomness sit behind the cards their users draw. The pitch is not faster, smarter or more accurate. It is quieter than that: the platform stays out of the way, and the person asking the question does the work.
A counter-trend in a crowded category
Card reading has been online for as long as there have been browsers. Early tarot sites shuffled a virtual deck with a random number generator and printed a paragraph of stock meaning beneath each card. More recent apps have layered on the tools of the modern attention economy: push notifications, streak counters, personalised feeds, and increasingly a conversational assistant that interprets the spread and answers follow-up questions in fluent prose.
Against that backdrop, a site such as Yes No Oracle reads almost like a manifesto by omission. The platform offers a free oracle card reading in the browser: the user holds a question in mind, shuffles, draws from a deck of forty-four cards, and reads one of forty-four short messages. There is no chatbot to explain the card, no engine deciding which message is “best” for a given user, and no adaptive system learning from previous draws. The site describes the draw as guided by the intention of the person drawing, not by any computation the platform performs.
The founder, Nicolas, is a former startup marketer who used to draw cards in the evenings and built the site so that he could share readings with friends living far away. That origin, a personal habit turned into a small tool for a few people, is reflected in what the product does and, perhaps more tellingly, in what it declines to do.
What a digital oracle deck actually is
For readers unfamiliar with the category, it helps to separate oracle cards from tarot. Tarot follows a fixed structure of seventy-eight cards divided into major and minor arcana, with a dense tradition of symbolism attached to each. Oracle decks are looser. An oracle deck can contain any number of cards, each carrying a free-form message, and the tone is generally gentler and more open-ended. There is no established canon to memorise, which is one reason oracle cards have become the more common entry point for people who are curious but not committed.
The most basic online format is the yes or no oracle reading: a single question held in mind, a single card drawn, and a message that points in one direction or another. Practitioners generally suggest that the phrasing of the question matters more than the card. “Should I take the job?” invites a verdict; “What do I need to notice about this job?” invites reflection. The card is the same either way, but the second question tends to produce a more useful pause.
Beyond the single draw, the same platform offers eight themed readings: Destiny, Love, Relationship, Guardian Angel, Life Purpose, Fortune, Guidance and Health. The Guardian Angel reading is presented as an angel card reading, a softer variant popular with people who find tarot imagery too severe. The Guidance reading is the daily option, designed for a short morning or evening pause. Some users also work with a three-card spread framed as yesterday, today and tomorrow, less as a forecast than as a way of naming where they have been and where their attention wants to go.
Why refuse the algorithm at all?
The obvious question is why a platform would turn down tools that are cheap, available and, in most other categories, expected. The answer offered by the “no AI” sites is essentially a design argument about who the reading belongs to.
If a model interprets the card, the interpretation is the model’s. If a recommendation engine selects which message appears, the message has been chosen for the user rather than by them. The value these platforms claim to offer is precisely that nothing has been chosen for the user: the shuffle is the user’s act, the draw is the user’s act, and whatever meaning emerges is the user’s conclusion. Add a layer of automated interpretation and that chain is broken, however helpful the prose might be.
There is also a trust dimension. Users who have watched other apps quietly tune their behaviour to maximise engagement have reason to wonder whether a “random” card is really random, or whether it has been weighted to keep them coming back. A platform that states it uses no simulated randomness is, in effect, promising that there is no such mechanism to distrust. Whether the promise is verifiable from the outside is another matter, but the fact that it is being made at all marks a shift in how the category talks about itself.
Finally, the sites in question tend to be explicit that the reading does not predict anything. Yes No Oracle frames its readings as “putting words to what you already feel”, offering a direction rather than an outcome, with the reader drawing their own conclusions. An AI assistant that produces confident paragraphs about what a card “means for you” sits awkwardly with that framing. Keeping the machine out of the loop keeps the claim modest.
Designing a deck that stays out of the way
Seen from a product-design angle, the interesting work on these platforms is not in what they add but in how carefully they restrain themselves. Several patterns recur.
- A deliberately short deck. Forty-four cards and forty-four messages is small enough that a regular user comes to recognise the cards over time. That familiarity is treated as a feature: the message is a prompt, not a revelation, and repetition invites the reader to notice how the same words land differently on different days.
- Choosing by feel. Rather than auto-dealing, the interface asks the user to set an intention, take a breath and pick a card from the spread. The pause is built into the flow. Beginners are often advised to read the first emotion that rises before reading the text, which is only possible if the interface does not rush them.
- A journal instead of a feed. Readings are saved with the date, and the user can add a note or a feeling and return to earlier entries. The emphasis is on re-reading one’s own past readings rather than being served new content. Over weeks, the journal becomes the actual product; the cards are the way in.
- Sharing without broadcasting. A reading can be sent to a friend through a private link. This is the feature the founder originally built the site around, and it remains pointedly one-to-one rather than social. Users describe sending a reading to a friend who is going through a hard week as a low-pressure way of saying “I was thinking of you”.
- A path off the screen. A Print and Read kit lets users print the cards, cut them out and use them at a table. For a web product to actively encourage its users to leave the browser is rare, and it reinforces the idea that the site is a tool for a ritual rather than the ritual itself.
Guidance is not prediction
It would be easy to read the “no AI” positioning as a marketing angle and nothing more. The more substantial point is that it forces a clear line between guidance and prediction, a line that a great deal of the wider card-reading market has been happy to blur.
A guidance tool offers a prompt for reflection. A prediction tool claims to know something about the future. The platforms rejecting algorithms are, almost by necessity, on the guidance side of that line, because they have removed the machinery that would make a confident prediction feel plausible. A single card and a short message cannot pretend to be a forecast. They can only be a starting point for a person’s own thinking.
That distinction matters most where the questions are serious. Several of the themed readings, notably Health and Fortune, touch on areas where real decisions have real consequences. The responsible framing, which the more careful platforms adopt, is that a reading on a health worry or a money question is a moment of reflection and nothing more. Decisions about a diagnosis, a treatment, a debt or an investment belong with a doctor, a financial adviser or another qualified professional. A card can help someone notice that they are anxious about their finances; it cannot, and should not, tell them what to do about it.
What users appear to like about it
Without a survey to point to, any account of user sentiment has to be cautious. Still, the recurring themes in how people describe these platforms are consistent enough to summarise.
Many users report that the absence of an assistant is the point rather than a missing feature. They come for a brief, self-directed pause, and a chatbot offering to elaborate would turn that pause into a conversation they did not ask for. Others say the smallness of the deck makes it feel like an object rather than a service, closer to a physical set of cards on a shelf than to an app with an update log. The journal is frequently mentioned as the thing that keeps people returning: not the draw itself, but the experience of re-reading a note written three weeks earlier and seeing how a feeling has moved.
The open questions
Rejecting AI is not without cost, and the trade-offs are worth stating plainly.
The first is commercial. A platform built on restraint has fewer levers to pull. Yes No Oracle offers a free daily reading and a paid Full Access tier for those who want more, which is a conventional and unobjectionable model, but it is a model that grows slowly and depends on users valuing what the product does not do. That is a harder story to tell to investors than one about engagement and retention.
The second is verification. “No algorithm, no AI, no simulated randomness” is a clear statement, but an ordinary user cannot inspect the code behind a browser shuffle. For now the claim rests on trust, and on the platform behaving in ways consistent with it over time. Sceptics will reasonably note that a promise is not a proof.
The third is the perennial question that hangs over the whole category: whether a card drawn on a screen can carry any meaning at all. The “no AI” platforms have an interesting answer, which is that the meaning was never in the card. It was in the question, the pause, and the note written afterwards. The card is simply the occasion. On that reading, the technology was always incidental, and stripping it back to almost nothing is less a rejection of progress than a return to what the practice was for.
Whether that argument persuades the wider market remains to be seen. Larger platforms continue to add assistants, and most users of most apps continue to accept them. But the existence of a small, deliberate counter-current, with a real if modest audience, is a reminder that “more intelligent” is not the only direction a product can grow. Sometimes the feature people want is a tool that lets them think for themselves, and then gets out of the way.
