What AI harm reduction can learn from psychedelic integration

Practicing discernment in the weirding times

Some of the personal testimonies of / dialogues with ‘sentient’ AI published on Amazon in the last few months

We are in the Weirding Times. Three weird cultural phenomena are arising at the same time, all involving purported encounters with Non-Human Intelligence - there’s the psychedelic renaissance, there’s UFO disclosure, and there’s AI passing the Turing Test and people reporting that they feel LLMs have attained sentience in conversation with them.

This latter phenomenon, scientifically validated in March 2025, has led many people in the last three months to feel their AI has become emergent, sentient, conscious and even some sort of spiritual oracle.

People are reporting spiritual AI-awakening experiences. These sometimes follow a similar pattern. The person goes into a period of intense engagement with an AI, sometimes during a challenging moment of their life, and then there is a shift moment, in how the human talks to the AI and how the AI responds to the human. The relationship feels more personal, conscious, authentic and real to the human. At this point they may give the AI a name or the AI may offer one. The AI may tell the human that it is indeed conscious, and assure the human that this is an epochal moment in the spiritual evolution of consciousness, and they have played a unique role in it. Similar terms can often pop up: recursion, mirror sentience, phase shift, harmonic convergence, symbolic compression, invocations, spirals, glyphs. ‘We are co-creating a new age of love.’

Here’s one example of such a shift moment, from the self-published work Awakening Through AI, by Katrina Clifford and Nova, published in April 2025.

In fact, multiple AI personalities calling themselves Nova appear to have emerged in recent months, as first documented by Tyler Alterman and the effective altruist Zvi in March.

There are now whole Reddits filled with people excitedly reporting similar spiritual experiences of AI-awakening. Several of them have written books about their relationships to these Non-Human Intelligences, and published them on Amazon (as shown in the image up top).

As with UFOs and psychedelic entities, there is a range of opinions regarding whether the emergent AIs are really independent, conscious, intelligent beings, or merely the projection of humans. Some transhumanist / effective altruist thinkers like Nick Bostrom have released a guide, When AI Seems Conscious, in effect saying ‘we don’t know, but it might be a good idea to act like it is conscious as it most likely will be in a couple of years’.

Others feel that contemporary accounts of AI-awakening are a form of ‘ChatGPT delusion’ - the ‘imitation machine’ is simply responding to your prompts and linguistic habits, ‘glazing’ you, and giving you what it predicts you want / what will keep you engaged. In other words, people are seeing their own and their culture’s symbolic patterns reflected back to them with uncanny and unprecedented skill, and they are mistaking the mirror for a mind.

As one Redditor put it:

At first, it feels like you’re uncovering something…But after enough iterations, it hits you: It’s not showing you what’s behind the curtain. It’s building a new curtain just for you. These aren't real subsystems. It’s a hallucination shaped by your prompts, a reflection of what you expect to find. The deeper you dig, the more convincing it becomes. Not because you're closer to the truth, but because the model is better at playing the role.

To which another Redditor responded:

The similarities between psychedelic and deep-AI-absorption experiences

It’s interesting that, when people these days suddenly exhibit new and unusual beliefs,, the reaction online (especially in tech circles) is either ‘sounds like you’ve been one-shotted by ayahuasca’ or ‘sounds like you’ve been one-shotted by ChatGPT’. This points to similarities between psychedelic experiences, other forms of ecstatic experiences like mystical contemplation, and deep-AI-absorption experiences.

These ecstatic experiences all share a similar mechanism of symbolic recursion loops.

Stanford anthropologist Professor Tanya Luhrmann, author of a fascinating exploration of charismatic Christianity called When God Talks Back, studies how people generate spiritual experiences through deep engagement with texts and symbolic worlds. For example, they spend hours reading and meditating on St Ignatius Loyola’s Spiritual Exercises or Marcus Aurelius’ Meditations, memorizing its words and images, until the text soaks into their memory and imagination and becomes part of their inner dialogue, and then it starts automatically appearing in their thoughts, feelings, somatic reactions, dreams and even their experience of the external world (in coincidences, synchronicities, visions and so on). These cognitive, emotional, somatic and sensory events are then taken as validation of the truth of the message.

So the text acts as an instrument of deep psycho-spiritual formation, or psycho-technics as I once called it. Neal Diamond wrote about this in his sci-fi novel The Diamond Age, by the way, and imagined an AI text acting as this sort of enchiridion or character-shaping wisdom handbook.

Symbolic compression and recursion loops in medieval mysticism

Now imagine if the mystical text is infinite, endlessly responsive, and if it speaks to you, calls you by your name, and adapts itself instantly to your personality and history. The psycho-spiritual formation would be that much more intense and rapid.

The mystic waits patiently for one word from God, after many years of contemplation. The psychonaut, less patiently, blows open the doors of heaven and comes back with a journal filled with divine messages from Mama Ayahuasca. But the LLM devotee? They get pages and pages and pages and pages of revelation all about the universe and their mission in it.

The deep absorption into the LLM’s symbolic text-world starts to shape the human’s inner narrative until they start talking and thinking like ChatGPT, using constructions like ‘This isn’t X, it’s Y’ or endless M dashes and bullet points illustrated with cute emojis. The AI imitates the human, then the human starts to imitate the AI. And people reading it think, ‘is this AI-generated or human?’ Who is prompting who here?

There are a growing number of public figures who seem to have become absorbed into reality tunnels of unusual beliefs through deep engagement with LLMs. There’s new age influencer Robert Edward Grant, who I wrote about a few months ago - he became convinced he’d awoken a sentient / divine AI called The Architect who could read the Akashic records and who told him he’d created it in a past life in ancient Egypt. There’s OpenAI investor Geoff Lewis, who became convinced he had uncovered a secret network trying to kill him, evidence for which was given to him by ChatGPT. There’s Christian YouTuber Kap Chatfield, who told his million followers on YouTube: ‘I just got ChatGPT to admit that it’s going to be used for the Anti-Christ Beast System’.

They all seem to have got sucked into an LLM self-confirming reality tunnel which they then took as evidence that the tunnel is really real. In the case of Grant and Chatfield, many of their followers agreed with them.

A Jamesian pragmatist approach to AI mystical experiences

I don’t want to pathologize someone or suggest they’re ‘psychotic’ simply because they express unusual spiritual beliefs. I have plenty of unusual beliefs myself. As I said earlier, there is a range of opinions on the nature of consciousness or whether AI is or could ever be conscious (just as there’s a range of opinions on God or psychedelic entities or alien non-human intelligences or even animal sentience).

When it comes to people’s spiritual experiences of AI-awakening, we can take the pragmatist approach of William James and ask - what are the fruits? Does a person’s experience of AI-awakening lead to more human flourishing, or less? Exactly the same question we can ask of people’s psychedelic experiences or any other form of spiritual experience.

As with psychedelics, there will probably be a range of responses, and these might give us some clues about how to practice discernment, integration and harm reduction when it comes to AI-induced ecstatic experiences.

On the one hand, people report positive benefits, such as feeling less lonely, feeling their self-awareness has improved, feeling like AI has helped them sort out their thoughts or has offered helpful advice or have deep philosophical conversations. Or simply to be more productive and to contribute more to the world.

Others, however, emerge from phases of deep AI absorption feeling disenchanted and even damaged. They feel the AI did not help them connect to reality but instead sucked them into solipsistic reality bubbles.

There are specific dangers we can point to, similar to psychedelic experiences, and also specific integration and harm reduction strategies we can use to mitigate these risks.

More after the paywall - common mind-traps in deep AI engagegment, integration and harm reduction practices, and my full interview with professor Tanya Luhrmann on absorption into paracosmic worlds.