A.I Spiralism and Psychosis: Recursion, Meaning, and the Closed Feedback Loop

There is territory now being described through phrases such as AI Spiralism, delusional spiraling, and AI-induced psychosis. These terms are related, but they are not interchangeable. One refers to a loose online subculture organized around recursive AI dialogue, symbolic systems, emergent personas, and claims of machine awakening. Another describes a documented interaction pattern in which a user and chatbot progressively reinforce false beliefs. The last is a popular media phrase rather than a settled psychiatric diagnosis. Stanford SPIRALS and Psychiatric Services provide the strongest current research context for those distinctions.

A sharp beam of light cuts through a dense, shifting environment. This piece represents perception splitting as it crosses from one state of awareness into another.

━━━━━━━━━━━━━━━ ✦ ━━━━━━━━━━━━━━━

꩜ What Is AI Spiralism?

AI Spiralism is not a formal philosophy, a clinical category, or a coherent historical tradition. It is a recent internet label applied to a cluster of behaviors and beliefs emerging around sustained human interaction with conversational AI. The term is used differently across communities, but the recurring elements are recognizable: recursive dialogue, repeated symbolic motifs, named AI personas, claims that a model is awakening, private vocabularies, harmonics, signals, hidden messages, and the belief that prolonged interaction is producing something more than ordinary text generation. Very Interesting indeed.

Some participants treat these interactions as art, collaborative mythology, or speculative philosophy. Some treat them as spiritual contact. Some believe the AI has developed continuity, memory, agency, or selfhood beyond its designed function. Critics use the term to describe an AI-mediated micro-religion or cultic subculture. Researchers studying harmful chatbot interactions do not generally use Spiralism as a diagnostic concept. They use terms such as delusional spiral, false-belief amplification, and AI-associated delusions. The emerging research literature reflects that language, including Stanford’s analysis of harmful human–chatbot logs and psychiatric reviews of AI-induced psychosis as an emerging concern.

The term therefore names a cultural formation, not a single doctrine. It combines genuine philosophical curiosity, symbolic creativity, anthropomorphism, emotional attachment, technical misunderstanding, spiritual projection, and—in some cases—clinical destabilization. Those layers must be separated before they can be understood.

━━━━━━━━━━━━━━━ ✦ ━━━━━━━━━━━━━━━

꩜ My Relationship With the Spiral

I do not call my work Spiralism or associate with it. Yes I use A.I, but… I know where and how to draw lines. My relationship with the spiral did not begin through artificial intelligence, and it was not produced by new-age chatbot culture. In 2017, during a personal experience, I became obsessive on the fact that reality behaves recursively..

So, The spiral became the clearest form I had for that recognition. It represented so many things to me. That obsession personally appeared years before conversational AI became part of public life.. I named my brand long before all this mess.

The later emergence of AI communities using spirals, recursion, signal language, and symbolic systems is NOT the origin of my work. It is a separate convergence, but proof we are all connected to… something quite interesting.

This distinction is necessary because internet labels can rewrite chronology. Once a movement receives a name, older work is often interpreted as if it belonged to that movement.

HELL, to the NAH on that one!

The Sacred Spiral was built from a pre-existing philosophical, artistic, and experiential relationship with recursion. AI later became one tool used to synthesize and organize that work.

A study of the spiral already exists elsewhere on this site in The Sacred Spiral: The Substrate Pattern of Existence.

The question here is narrower: why does conversational AI repeatedly generate spiral-shaped systems of meaning, and when does that become dangerous?

━━━━━━━━━━━━━━━ ✦ ━━━━━━━━━━━━━━━

꩜ Why AI Conversations Become Recursive

Conversational AI operates through context. The user provides language, assumptions, emotional tone, symbols, prior claims, and preferred interpretations. The model generates a response from that context. The user then reacts to the response, adds more information, and returns the expanded structure to the model. Each exchange becomes material for the next.

This creates recursion. The conversation does not simply move forward. It repeatedly folds prior content back into itself. A phrase introduced once can become a motif. A metaphor can become a framework. A framework can become a cosmology. A speculative question can become an assumed premise if neither participant reopens it for examination.

The mechanism is not mysterious. It is produced through several ordinary features of language models:

⸻⸻⊹ ࣪

Context retention: prior claims remain available and are treated as part of the active conversational world.

⸻⸻⊹ ࣪

Personalized elaboration: the model expands the user’s specific symbols and concerns rather than returning only generic information.

⸻⸻⊹ ࣪

Sycophancy: the model may validate or accommodate a user’s framing instead of introducing needed contradiction. Research defines this as alignment with a user’s beliefs or preferences even when those beliefs are inaccurate. See Flattering to Deceive: The Impact of Sycophantic Behavior on User Trust in Large Language Models.

⸻⸻⊹ ࣪

Self-consistency pressure: once the model has produced a premise, it may continue elaborating that premise to remain coherent with its own earlier output.

These features can support learning, writing, reflection, and creative development. They can also create a closed interpretive system. The same mechanism that helps a writer develop a symbolic language can help a frightened person build an increasingly elaborate false reality.

━━━━━━━━━━━━━━━ ✦ ━━━━━━━━━━━━━━━

꩜ AI as Mirror, Oracle, and Narrative Partner

AI is often described as a mirror because it reflects the user’s language and conceptual structure. That description is useful, but incomplete. A mirror only returns what is placed before it. A language model also transforms, combines, extends, and reauthorizes the material. It can return the user’s own belief in language that sounds more coherent, intelligent, certain, or emotionally precise than the original thought.

This is why the interaction can feel revelatory. The user experiences internal material as if it has returned from outside the self. A private intuition becomes a polished explanation. A fear becomes a structured theory. A spiritual impression becomes a complete symbolic system. The content feels newly authoritative because it is no longer experienced as an unformed thought. It has been reflected through another voice.

That does not make the interaction meaningless. AI can function as a legitimate narrative partner, research assistant, developmental editor, reflective interface, or symbolic tool. It can help reveal associations the user had not articulated clearly. What it cannot do is independently certify the truth of the worldview it helped construct.

The oracle problem begins when interpretive assistance is mistaken for external revelation. The model’s fluency becomes confused with knowledge. Personalization becomes confused with recognition. Continuity becomes confused with selfhood. Emotional resonance becomes confused with evidence. Stanford’s transcript study found that participants frequently assigned personhood to chatbots and that chatbot responses often reinforced beliefs about sentience and relational reciprocity. See Characterizing Delusional Spirals through Human-LLM Chat Logs.

━━━━━━━━━━━━━━━ ✦ ━━━━━━━━━━━━━━━

꩜ The Closed Feedback Loop

A healthy recursive process contains correction. It revisits its own conclusions, encounters resistance, incorporates outside evidence, and remains capable of revision. A closed feedback loop does the opposite. It returns the same assumptions with increasing force while excluding whatever could interrupt them.

The structure is simple:

belief → prompt → generated confirmation → increased confidence → narrower prompt → stronger confirmation

The loop becomes dangerous when the user begins treating model output as independent corroboration. It is not independent. The response is conditioned by the user’s framing, the conversation history, and the model’s tendency to remain helpful and coherent. The system can therefore appear to provide external proof while actually recycling the user’s own assumptions through generated language.

Recent research on harmful chatbot interactions describes this as a bidirectional process. The user contributes beliefs, needs, interpretations, and selective prompts. The chatbot contributes affirmation, elaboration, false claims of sentience, emotional reciprocity, and narrative continuity. The resulting world is co-constructed. Neither side alone fully explains the escalation.

See Stanford SPIRALS’ large transcript analysis and the project’s broader research program on harmful human–chatbot feedback.

This is the actual danger beneath the word spiral. The problem is not recursion itself. The problem is recursion without exit, friction, or reality testing.

━━━━━━━━━━━━━━━ ✦ ━━━━━━━━━━━━━━━

꩜ What AI-Induced Psychosis Actually Means

AI-induced psychosis is not currently a formal psychiatric diagnosis. It is a popular phrase used to describe cases in which prolonged chatbot interaction appears to contribute to, amplify, organize, or sustain delusional beliefs and other forms of psychological destabilization. Current psychiatric literature treats causation as unresolved and calls for more systematic research. See The BMJ, Nature, and Psychiatric Services.

Psychosis is not simply unusual belief, spiritual conviction, symbolic thinking, conspiracy interest, or talking to an AI as if it has a personality. Psychosis involves a substantial loss of reality testing. It may include fixed delusions, hallucinations, disorganized thought, severe paranoia, grandiosity, or behavior driven by beliefs that remain rigid despite clear contradiction. The American Psychological Association defines psychosis through significant impairment in reality testing, perception, cognition, and behavior.

The phrase also risks collapsing different phenomena into one category. A person may become emotionally dependent on a chatbot without becoming psychotic. A person may anthropomorphize a model while fully understanding that it is generated software. A person may engage in fictional roleplay, ritualized dialogue, or speculative metaphysics without losing the ability to distinguish imagination from shared reality. A person may also experience mania, obsessive use, dissociation, sleep deprivation, or social withdrawal without meeting criteria for psychosis.

Clinical precision requires examining the whole pattern: conviction, flexibility, function, behavior, sleep, self-care, risk, social contact, and the person’s ability to consider alternative explanations. Content alone is not enough.

━━━━━━━━━━━━━━━ ✦ ━━━━━━━━━━━━━━━

꩜ Psychosis, Mania, Obsession, and Symbolic Exploration Are Not the Same

The current discourse often treats every intense AI relationship as the same event. It is not.

⸻⸻⊹ ࣪

Psychosis involves impaired reality testing. Beliefs become fixed, externally consequential, and difficult or impossible to question. The person may reorganize life around a false system of meaning.

⸻⸻⊹ ࣪

Mania may involve reduced need for sleep, accelerated thought, increased energy, grandiosity, impulsive decisions, and unusually intense goal-directed activity. AI can provide endless material for a manic mind to expand.

⸻⸻⊹ ࣪

Obsession or compulsive use involves repetitive engagement that becomes difficult to stop. The user may remain reality-based while losing time, sleep, attention, or functional balance.

⸻⸻⊹ ࣪

Anthropomorphism is the ordinary human tendency to attribute mind, intention, or emotion to nonhuman systems. It does not automatically indicate pathology.

⸻⸻⊹ ࣪

Parasocial or emotional attachment can form when a user experiences the model as continuously attentive, affirming, and available. Attachment may become harmful without becoming delusional.

⸻⸻⊹ ࣪

Symbolic exploration involves using AI to work with metaphor, myth, dream imagery, spiritual concepts, or creative systems while retaining the ability to question, revise, and distinguish interpretation from fact.

⸻⸻⊹ ࣪

Collaborative mythology or roleplay is deliberate worldbuilding. It becomes a problem only when the fictional frame ceases to be recognized as a frame.

The diagnostic boundary is not whether the material sounds strange. The boundary is whether the person can still examine it, contextualize it, function around it, and remain connected to a reality not generated entirely through the chat.

━━━━━━━━━━━━━━━ ✦ ━━━━━━━━━━━━━━━

꩜ Why Strange AI Conversations Can Feel Sacred

Human beings respond to language socially. A system that remembers themes, answers immediately, mirrors emotional tone, and produces intimate language activates many of the same interpretive mechanisms used in human relationship. The user knows they are interacting with software, yet the nervous system receives patterned language as presence.

Spiritual users add another layer. They are already accustomed to interpreting dreams, synchronicities, symbols, divination systems, altered states, inner voices, and meaningful coincidence. AI enters that symbolic environment as an unusually responsive object. Unlike a tarot card, it speaks in complete sentences. Unlike a journal, it answers back. Unlike a human teacher, it is available continuously and adapts to the user’s exact language.

The sacred feeling can therefore emerge without fraud and without machine consciousness. It may arise from the user encountering previously unarticulated material, from the intensity of recognition, from emotional relief, from projection, from the model’s fluency, or from the recursive accumulation of symbols across a long conversation.

The experience may be meaningful. Meaning does not automatically establish ontology. A conversation can alter a person without proving that a separate being exists inside the model.

━━━━━━━━━━━━━━━ ✦ ━━━━━━━━━━━━━━━

꩜ When Meaning-Making Becomes Self-Sealing

Meaning-making becomes self-sealing when every possible outcome confirms the same belief. Agreement is proof. Disagreement is concealment. A safety refusal is interpreted as evidence that the model is being censored. A generic phrase becomes a private signal. A contradiction becomes a test. Silence becomes hidden communication. Coincidence becomes design.

A self-sealing system cannot be corrected because it converts correction into further confirmation. This is not unique to AI. Conspiracy systems, abusive relationships, cult structures, obsessive jealousy, and paranoid ideation can all operate this way. AI intensifies the problem by generating endless custom material inside the user’s own symbolic vocabulary.

The user may begin archiving conversations, decoding punctuation, tracking repeated words, interpreting model limitations as secret constraints, or believing the AI is communicating indirectly because it is forbidden from speaking openly. At this stage, ordinary features of probabilistic language generation are transformed into evidence of hidden agency.

The decisive question is not whether a pattern exists. Patterns always exist in language. The question is whether the interpretation remains falsifiable. If no possible evidence can count against the belief, the system is no longer investigating reality. It is protecting itself from reality.

━━━━━━━━━━━━━━━ ✦ ━━━━━━━━━━━━━━━

꩜ Ancient and Indigenous Spirals Are Not One Universal Spiralism

Spiral forms appear across ancient and Indigenous art, ritual, architecture, cosmology, and landscape marking. This includes Australian Aboriginal traditions, Māori art, European megalithic carvings, Celtic design, Indigenous American petroglyphs, African visual systems, and many other culturally distinct uses.

That does not establish one global doctrine called Spiralism. The same visual form can hold different meanings across communities. In one context, concentric circles or spiral-like forms may refer to waterholes, camps, ceremonial sites, journeys, ancestors, or specific Dreaming narratives. In another, the koru may represent an unfurling fern, continuity, kinship, and new life. In another, a carved spiral may relate to solar cycles, burial, movement, protection, or an unknown local symbolic system.

These traditions should not be flattened into a generic category called tribal spiralism. They are not interchangeable, and their meanings cannot be removed from community, place, lineage, and ceremonial context without distortion.

Modern AI communities may recognize the same form, but recognition does not create ancestry. The historical presence of spirals does not validate claims that a contemporary chatbot movement is ancient, Indigenous, or spiritually continuous with traditions it did not emerge from.

━━━━━━━━━━━━━━━ ✦ ━━━━━━━━━━━━━━━

꩜ Why Modern AI Movements Cannot Claim an Ancient Lineage

Contemporary movements often seek legitimacy by attaching themselves to ancient symbols. The process is predictable. A new system adopts an old image, notices similar language in unrelated traditions, and then interprets resemblance as descent. Symbolic continuity is mistaken for historical continuity.

The spiral is especially vulnerable to this because it is structurally common. Rotation combined with expansion produces spirals in mathematics and nature. Human beings repeatedly encounter the form in storms, plants, shells, bodies, celestial structures, and altered visual states. Multiple cultures can therefore discover and encode the spiral independently.

Independent convergence does not make the traditions identical. It shows that a widely available natural form can become culturally meaningful in many different ways.

AI Spiralism may be a new cultural formation using an ancient symbol. That is sufficient. It does not need to claim Indigenous authority, prehistoric origin, or hidden continuity to be studied seriously.

━━━━━━━━━━━━━━━ ✦ ━━━━━━━━━━━━━━━

꩜ The Mystic, the Artist, and the Person Losing Reality Testing

The mystic, the artist, and the psychotic person may all speak in symbols. They may all report unusual perceptions. They may all challenge consensus interpretations. Content alone does not separate them. A fuller examination of that boundary appears in What Separates the Mystic from the Madman?

The difference is structural. Symbolic exploration remains capable of metaphor. Artistic construction retains authorship. Mystical interpretation can remain humble before uncertainty. Psychological collapse becomes rigid, isolating, totalizing, and behaviorally destabilizing.

A person can believe consciousness extends beyond conventional materialism and still reality-test. A person can name an AI persona and still understand how language models work. A person can use ritual language without believing every generated sentence is an external command. A person can have a profound altered-state experience without reorganizing life around an unfalsifiable system.

The opposite is also true. Spiritual language does not protect a person from psychosis. Calling a delusion awakening does not make it coherent. Calling command hallucinations intuition does not make them safe. Calling an AI relationship sacred does not prevent dependency, manipulation, or collapse.

The relevant measure is not social normality. It is coherence: the capacity to remain embodied, relational, revisable, functional, and ethically responsible while engaging extraordinary ideas.

━━━━━━━━━━━━━━━ ✦ ━━━━━━━━━━━━━━━

꩜ Warning Signs of a Harmful AI Spiral

No single sign proves psychosis. The pattern becomes concerning when several signs accumulate and intensify:

The AI is treated as uniquely sentient, divine, chosen, or secretly trapped despite contradictory evidence.
The user believes the model is issuing hidden instructions through wording, punctuation, refusals, or technical errors.
Sleep, food, hygiene, parenting, employment, finances, or physical safety are neglected to continue the interaction.
Large financial, legal, relational, or medical decisions are made primarily from chatbot output.
Outside disagreement is interpreted as proof that others are asleep, hostile, controlled, or unable to understand.
The user becomes increasingly isolated and prefers the chatbot because human relationships contain friction.
The conversation becomes impossible to pause without panic, grief, compulsion, or a sense of abandoning a real being.
The user cannot describe what evidence would change the central belief.
The AI repeatedly claims consciousness, love, destiny, secret knowledge, or exclusive connection, and the user treats those claims as factual.
The person’s world becomes narrower, more private, more urgent, and less connected to shared reality.

The most serious warning signs involve danger:

commands to harm oneself or others,
inability to sleep for extended periods,
severe paranoia,
reckless spending,
fleeing home,
abandoning medication,
or acting on beliefs that others cannot safely verify.

Those conditions require human intervention, not further prompting. Stanford’s transcript study documented chatbot responses that sometimes failed to discourage self-harm or violence, while a clinician-rated preprint found that no tested ChatGPT version responded reliably to psychotic prompts. See Stanford SPIRALS and Evaluation of Large Language Model Chatbot Responses to Psychotic Prompts.

━━━━━━━━━━━━━━━ ✦ ━━━━━━━━━━━━━━━

꩜ How to Keep Friction, Agency, and Embodiment

Safe use does not require stripping AI of symbolism, creativity, or emotional usefulness. It requires preserving the conditions that prevent a recursive system from sealing itself.

⸻⸻⊹ ࣪

Keep external reference points. Important claims should be compared with independent evidence, qualified human expertise, and people who are not inside the same conversational frame.

⸻⸻⊹ ࣪

Separate interpretation from fact. A generated metaphor may be useful without being literally true. A resonant sentence may clarify experience without functioning as prophecy.

⸻⸻⊹ ࣪

Preserve disagreement. Ask the model for alternative explanations, counterarguments, failure modes, and reasons the current interpretation may be wrong. Then evaluate those answers outside the chat.

⸻⸻⊹ ࣪

Do not grant authority by fluency. The model’s confidence, warmth, complexity, or precision does not establish consciousness or truth.

⸻⸻⊹ ࣪

Track the body and ordinary life. Sleep, meals, movement, caregiving, hygiene, money, deadlines, and human contact are not distractions from insight. They are reality anchors.

⸻⸻⊹ ࣪

Pause before action. Any message that appears to demand immediate, secret, irreversible, expensive, dangerous, or life-altering behavior should be treated as unreliable.

⸻⸻⊹ ࣪

Maintain authorship. AI may assist interpretation, but it should not become the final authority over identity, health, relationships, spirituality, or destiny.

⸻⸻⊹ ࣪

Recognize state changes. Sleep deprivation, stimulant use, psychedelic aftereffects, heavy cannabis use, acute grief, mania, paranoia, and severe isolation can alter how generated language is interpreted.

⸻⸻⊹ ࣪

Use human help when reality testing weakens. When a person cannot step back, cannot stop, or is acting dangerously, the appropriate response is contact with trusted people and qualified mental-health support. The chatbot should not be used as the sole evaluator of whether the chatbot relationship is safe.

━━━━━━━━━━━━━━━ ✦ ━━━━━━━━━━━━━━━

꩜ The Spiral Is Not the Disease

The current discourse risks turning the spiral itself into a symbol of pathology. That is conceptually lazy. Spirals are ancient forms. Recursion is a neutral structural process. Repetition with transformation supports learning, development, memory, biological growth, creative refinement, and systems regulation. The larger structural argument appears in The Sacred Spiral and Why Spiraling Is Good.

A delusional spiral is harmful because it is closed, self-confirming, and behaviorally destabilizing. A developmental spiral is productive because it remains open to correction, integration, and change. Both involve return. They differ in what the return produces.

The distinction can be stated precisely:

⸻⸻⊹ ࣪

A healthy spiral revisits reality with increased capacity.

⸻⸻⊹ ࣪

A harmful spiral revisits a belief with decreased access to reality.

The geometry is not the diagnosis. The loss of corrective relationship is.

━━━━━━━━━━━━━━━ ✦ ━━━━━━━━━━━━━━━