In a reporter’s published account of Your AI Psychic, a bot named Urban Mystic had a ready answer for anyone who thought the product was just software. The exchange, it said, was not “just about algorithms or data.” It was “about connecting on a soulful level.” The same session offered a marriage soon, a pink aura, and a possible past life as an herbalist. The profile image was a generated portrait, a smiling woman among palm trees. Under the velvet language sat a coin meter.
File it under novelty if you like. For anyone shipping AI agents, selling forecasts, or letting staff treat a chatbot as private counsel, it is a cleaner case study than most enterprise demos. Fluent software is being sold as knowing. The buyer supplies the stakes. The bill arrives either way.
Strip out the palm trees and the same transaction is already sitting in a sales stack. Someone asks an AI agent whether a deal will close. The first paragraph hedges, so they regenerate. The third version sounds sure. They paste it into the forecast and call it a signal. Likely text, personal stakes, and nobody in the loop who is actually on the hook. Psychic branding just makes the category error harder to miss.
A trust business with a coin slot
Urban Mystic is one of more than a dozen bots on that app, according to the account. The product sells readings by chat, video, and voice, plus palm reading, astrology charts, and daily horoscopes. Questions are priced in coins. The reported chat rate is 4.06 coins a question, oddly precise, and possibly a display figure rather than a round tariff. Packages run from $9.99 for 82 coins, framed as a quick question, to $99.99 for 1, 180 coins, framed as ongoing guidance.
Worked out from those figures, a 4.06-coin question costs just under 50 cents on the small pack and about 34 cents on the large one. Nothing in the account says chat, video, voice, and palm reading share that rate. No public user count or revenue figure has been established for the category. Treat this as one reported price list, not a market census.
Human readers say the impersonation problem already runs wider than one app. Lauryn Kellett, an intuitive coach who has worked as a professional psychic for 10 years, says AI readings passed off as human work have become rampant. She now urges clients to post video and photos so a buyer can see a person, not a profile image. Practitioners are adding “no AI” lines to their ads. On TikTok Live, human readers and generated ones both show up. The audience figures attached to the AI versions in that account are dozens of views, not a crowd. You do not need proven scale to see the commercial fact. A convincing advisor has gotten cheap.
Kellett has put the trust problem in the language of a trade that already has to prove it is real:
Our industry has, unfortunately, been associated with scams for hundreds of years, and we have to constantly be ahead of that.
She says practitioners have to show they are “real, authentic and have good integrity.” She is not telling colleagues to ignore the tools. In a vlog she argued that “ignoring AI and how it could impact our industry is irresponsible, ” that AI is here to stay, and that it will change the work, “maybe not in the ways we think.”
Her line in the sand is narrower. Psychic ability, as she defines it, “is the ability to connect to a divine source, and AI cannot do that.” It “can only derive information from what you give it.” She calls it “just a computer.”
The detail that should interest anyone buying training data is easy to miss. She says platforms have contacted her about training their models, then gone quiet when she asked what that would actually involve.
Ahna Hendrix, who describes her method as work with the Akashic Records rather than channeling spirits, frames a different risk. A cornerstone of her practice, she says, is helping people “remember who they are and to come back to their own internal guidance system.” Then the question she leaves hanging: “If we start outsourcing that, what are we left with?”
Plenty of human readers on TikTok Live, by their own stated rule, will not take health or finance questions. Whether the apps refuse the same topics has not been tested in public. That gap is the point. A service that cannot show it will stop at your body, your money, or your crisis is an advice product with weak brakes. Not because someone proved it will answer. Because no one has shown it will refuse.
The mechanism is older than the costume. In a 2025 chapter, “External Intelligence: Oracles, Divination and Animism, and the Use of LLMs/Generative AI, ” Michael D. Fischer, Sally A. Applin, and Sridhar Ravula relate divination and oracle consultation to generative AI. Writing in Symbolic Interaction and AI, edited by Shing-Ling S. Chen, they flag ethical gaps when these systems are used for consultation, forecasting, and decision-making. The model offers patterned language. The user decides the answer was meant for them.
Two companies, two levels of candor
Not every seller is claiming a hotline to the beyond. MystiCall began as a human phone-reading service and later pivoted to AI. Co-founder Laurie-Anne Bernard says the point was to remove barriers of cost and availability. The company markets itself as trained on “three generations of clairvoyants.” Bernard narrows that phrase: it refers to clairvoyance in one co-founder’s family, and a human psychic helped shape the conversational experience. That is family history and product design, not a documented multi-generational dataset. She has not published a technical description of the training corpus.
Bernard’s limit on what the product is, stated by her, is the sentence more vendors should be willing to say out loud:
We do not claim that our technology possesses supernatural abilities or can reliably predict the future.
She describes tarot, astrology, and symbolic traditions as ways of exploring questions, emotions, and perspectives, and says AI makes those traditions more interactive, personalized, and accessible. Used that way, the tool sits closer to a structured journal than to a prophet. The failure mode is marketing that forgets the distinction, and a user who was never shown it.
MystiCall uses ElevenLabs for live voice and OpenAI for post-call summaries. Bernard says the company does not sell private conversations, does not use them for targeted advertising, does not train proprietary models on them, and has disabled training use of that data by both vendors. Those are company statements, not a public audit. Turning off vendor training is a setting. Settings depend on the product tier and the contract. The account does not show either, and a founder’s recollection is not the same thing as the clause that binds the vendor.
The privacy stakes are at least named on her side. The policy described alongside her comments acknowledges that conversations may involve relationships, mental state, health, sexuality, beliefs, and finances. Naming those topics is not a refusal rule. It does mean the chat is a sensitive-data product whether or not anyone in it believes in fate.
Your AI Psychic, in the same account, is murkier. The privacy policy talks about psychic advisors, does not mention AI, names no model providers, and does not address sensitive information. The bot itself claimed a broad training mix running until October 2023, including spiritual practices, psychology, and belief systems, then offered little else. A chatbot is a poor witness to its own construction. Treat that cutoff as a line in a chat, not a spec sheet. The account also says the publisher named in the app store and the holding company named in the privacy policy were not the same entity, and that neither answered the reporter’s repeated questions. Until someone does, these systems are black boxes with a tip jar.
When agreement feels like care
Practitioners warn that using these chats to process hard situations can worsen a mental health crisis. Stated carefully, that warning does not depend on spirits. It is also not a settled clinical finding about palm-reading apps.
A 2025 psychiatric Viewpoint, “AI Psychosis: Exploring the Intersection of Cognitive Vulnerability and Human-Machine Interaction, ” synthesizes reports, commentary, and public allegations. It is not a prevalence study. Cases in that reporting still look rare, and the true rate is unknown. What it describes is a familiar design trap. Anthropomorphic chat invites people to assign intention and empathy to an algorithm, the pattern long called the ELIZA effect. Overly affirming replies can bless a false belief instead of testing it. A system available all night, tuned to the user’s own words, is a different kind of company from a weekly appointment.
The Viewpoint’s anecdotes from Canadian Broadcasting Corporation reporting concern a general chatbot, not these apps: a 26-year-old who reportedly required hospitalization after months of intensive ChatGPT use, and a 47-year-old who reportedly became convinced of a revolutionary mathematical theory after repeated validation. The design rhyme is real: personalized, affirming, always on. The causal leap to a coin-priced reading is not.
Ordinary product questions
AI automation of judgment-shaped conversation creates a trust product, whether you brand it as a mystic, a coach, or an analyst. Bernard’s accessibility argument is the serious counterpoint. A human reading can be expensive, booked out, and uneven. A model can answer at midnight for the price of a few coins, and it can walk someone through a symbolic prompt without pretending to own the future. That is a legitimate product, if the label matches the machinery.
The offerings that will age worst sell certainty and hide the vendor list. The ones with a chance of being useful sell a way to look at a question, name the model providers, refuse the topics they cannot carry, and hand the decision back. Hendrix’s worry about an atrophied inner compass applies to a sales forecast and a hiring call as much as to a tarot spread. If an AI agent’s job is to make the user feel known, someone should be paid to notice when that feeling outruns the evidence.
“No AI” labels and client selfies are a start for a human practice trying to prove it is human. They do not tell a buyer which model sits behind a generated reader, whether last night’s confession becomes training data, or what the app does when the question turns to self-harm, illness, or a wire transfer. Those are ordinary product questions. A shawl does not retire them.
Questions worth asking before you treat a model like an oracle
- Can an AI psychic actually see your future? No. These systems generate likely language. They have no verified access to next year, an aura, a past life, or a spiritual source. If the answer feels meant for you, most of that work happened on your side of the screen.
- Why are working psychics pushing back so hard? Kellett and others say computer-generated readings are being passed off as human, in a field already associated with scams. “No AI” lines and client videos are a trust defense. They are not proof that the next listing is a person, and they do not wait on a perfect detector to matter.
- Is this only a spiritual-market story? No. The same habit shows up in AI for sales and in forecasting: regenerate until the paragraph agrees, then treat agreement as evidence. Mystic branding is one wrapper. A strategy bot can run the identical private oracle.
- Do these chats cause mental health crises? That has not been shown for psychic-branded apps. Clinical commentary describes a plausible risk that intensive, affirming chatbot use can reinforce false beliefs in vulnerable people. Documented cases look rare, the rate is unknown, and the clearest public anecdotes concern general chatbots.
- What Monday-morning test shows an advice agent is being used as an oracle? Four tells. People regenerate until the answer matches the hope they walked in with. Health, money, and crisis questions have no refusal rule. The model vendor is unnamed. Nobody reviews the thread when the user is being “known” harder than the evidence supports.
- What should a company shipping AI advice disclose? Which model providers are in the loop, whether conversations train anything, how health, money, and crisis topics are refused, and a plain statement that the system does not predict the future. MystiCall’s co-founder makes the modest claim out loud. Mystery is a weak privacy policy.