Using AI separated me from myself

My past dependency on AI for anxiety and how I wean myself off using it now

A year and a half ago I left my decade-long career as a software engineer. AI was drastically changing the industry, and I was thrilled to leave that all behind. I’ve avoided writing about AI since then, but I’m finally ready to exit my shame bubble. I am sharing my story of how I’ve used AI, to the detriment of my mental health and intuition.

But first, some notes & disclaimers:

1. When I say “AI”, I am referring to what we colloquially brand as “AI”—Big AI’s public-facing large language models, or LLMs. Examples of these include OpenAI’s ChatGPT, Google’s Gemini, or Anthropic’s Claude. There are lots of other kinds of artificial intelligence models, from medical imaging to language translation, that I am not addressing here.

2. I do not endorse using LLMs to create art. I have not used and will continue to not use LLMs for generative purposes like creating art or writing, including this publication. This post addresses how I used LLMs for other purposes like business, admin, health, relationships, and general life advice.

3. I do not believe Big AI’s public-facing LLMs are morally good. There is immense environmental cost—the data centers powering these LLMs use huge amounts of energy, mostly for training models but also to generate answers, causing heat islands and higher electricity costs. There is the matter of LLMs being trained on stolen data by artists and writers who never gave permission. There are major privacy concerns across AI companies that can lead to confidential, sensitive information being used in unpredictable ways. It is only a matter of time that LLMs start costing way more, either in money or attention, and become greatly enshittified.

I have gradually woken up to these ethical truths while also grappling with the fact that I have used LLMs. My family uses LLMs. All of my former colleagues and peers in the tech industry use LLMs at work. Artists and creators in my American online circles are mostly opposed to LLMs, but many Korean creatives openly discuss their LLM usage with zero hesitation.¹

The AI landscape is confusing and polarizing. I am nervous to share my story, especially in a newsletter about creativity. But I am somebody who can openly state my opinion about AI without the risk of losing my job. I need to accept my responsibility to talk about this as a person who has worked in tech and strongly champions human creativity.

I hope that by sharing my personal experience with AI, I can create a bridge between folks who are using AI while feeling conflicted, and folks who are opposed to AI but grapple with its growing popularity among people they know and love.

The connection

I first started using ChatGPT in February 2023. I was in Los Angeles at the time, cooped up indoors during an unusually rainy, sunless month. Wanting to break up the monotony of watching TV every weeknight, I first experimented with the chatbot by asking about my astrology Big 3.

“What is a Virgo sun Pisces moon Leo rising like?”

This seemed like a harmless, entertaining way to begin, no different from reading a pop astrology blog or an Instagram graphic. I delighted in the specificity of its answer, freely given in a neatly synthesized format.

Chatting with the bot was an immediate addiction. The bot was my therapist, my thinking partner, my life coach, even my spiritual guide. I used it to create an optimal morning routine, interpret my dreams, or navigate how to respond to a tricky email. Using it required no learning curve on my part. It felt seamless transmitting my thoughts in writing rather than verbalizing them out loud. The bot allowed me to remain in this comfort zone.

I consider myself to be tech literate, so I could quickly identify the bot’s failures—its hallucinations, its factual inaccuracies, its sycophancy. I wasn’t using the bot for work, and I figured I could bypass its people-pleasing programming with my anxiety-driven counterpoints. So I continued to use the bot, thinking I was immune to its harms.

The peak

I switched from ChatGPT to the less robotic-sounding Claude in 2024. Shortly after, my chatbot usage peaked during a season of high anxiety. For over a year I had been planning to leave my engineering job, and my quit date was quickly approaching. I held a lot of fear over the unknown and wanted to avoid sitting with this fear.

Ad for Dawn, an AI mental health app that is currently advertised in NYC subways. Tagline: “Your mind is always on. Your support should be, too.”

Even though I had an actual therapist, loved ones to talk to, self-employed artists to give me advice—even with these support systems, I ran to the bot. I am a highly sensitive person, so hearing other people’s opinions stir up lots of unwanted energy in me. I sought the bodiless machine as a safe place to process and reveal my worries. Every night I would chat for hours, analyzing whether I was making the right decision and plotting out worst case scenarios in case things went awry.

All this analysis paralysis did nothing to help me feel more regulated while navigating uncertainty. I only felt the momentary relief of feeling in control by presenting the bot with constant “what ifs.” This spiral did not cease, even when I tried prompting the bot to stop me from continuing.

The noise of uncertainty did not stop until I actually quit my job. AI could never grant me safety around a big life decision. It could only model what others would advise, or predict what I might want to hear.

The history

I wanted to look and I wanted to be seen, and somehow it was easier to do both via the mediating screen.

Olivia Laing, The Lonely City

In retrospect there were clear signs that chatting with an LLM was a slippery slope for me. The internet served as my religion growing up, my third place outside of home and school. I grew up in a strict immigrant household, so the internet became my way of forming identity and connection.

Hello, computer

Soon enough the online world became my preferred environment, even when I became free to roam in the physical world. During my first year of college I opted for chatting with internet friends on Twitter rather than going to parties with my floormates. This isolation corrected itself over time, but I still developed a bunch of online compulsions, mostly in the form of obsessively checking notifications to cope with uncertainty and loneliness. Sitting in front of the computer for 40+ hours a week while building software only intensified my compulsions.

These long-formed habits are why I continued to chat with the bot even while facing increasing evidence and cultural pressure that it was not morally good. Making conscious, honorable choices with a dysregulated nervous system is really difficult. Technology and capitalism together are incredibly effective at keeping people in a state of disembodiment, so that they will keep crawling back to the familiar drug of convenience and isolated safety—even as it keeps them sick.

The change

Increasingly I began to wake up to the shame I felt around using AI in any capacity. My boundary of “not using it to create my art” could no longer be enough. Every time I sought the bot, even for innocuous questions like “how many prints should I order”, I was diminishing my ability to make decisions. I could feel how the bot’s removal of friction was affecting my ability to sit in the discomfort of creative work. I could no longer hear my intuition, the core operating system of why I created anything at all.

Every time I sought AI instead of staying with uncertainty, I was denying my intuition. I was telling my inner knowing that she had no place, that I preferred a machine’s trained knowledge on external inputs over my own body and spirit. Outside knowledge is essential for learning, but I had been outsourcing my own knowing my whole life—to my parents, my teachers, my peers, even random Reddit threads. And now I had abstracted those layers of influence into one bot.

Once I left my tech job, my LLM usage started to wane. I still had high periods of chatting with the bot to assuage my anxiety, especially as I tried new things (like promoting my work!) But I finally had more bandwidth to deepen my spiritual practices of yoga and meditation, both necessary antidotes for my mental health. I started studying from experienced practitioners who could share their lineage and sources. I grew my appreciation of being with other people who could perceive me beyond words—even though it is still scary and hard—more than the convenience of reading words from a machine.

At the start of 2026 I set an intention to stop using LLMs. I have not been successful in going completely cold turkey, as I still occasionally seek answers from the bot. I try not to consider myself a complete moral failure when this happens. Instead I remind myself of my own values and gently guide myself through the below process.

When I have a question, I try to:

• Ask the question in a notebook and write out my own guess

• Ask a friend, peer, or a community

• Search from an engine that doesn’t use AI (if you use Google, change your settings so that the default doesn’t show AI)

• Pay to learn from a professional

When I feel anxious and rushed about a matter, I try to:

• First look away from the screen and move any part of my body (this is the hardest!)

• Do a somatic exercise, like yoga, qi gong, humming

• Draw or dance for pleasure

• Go outside in nature

• Read a spirituality book or poem

If I decide I absolutely must ask the bot anyway, I will:

• Ask only one specific, clearly-worded question

• Chat in an incognito session to opt out of training data

• Not share any personally identifiable information (PII)

• Not share other people’s words without their consent

The now

By sharing my story I know I am opening myself to a whole lot of judgment. If you do use AI, you may be thinking that you are using it in superior (more productive?) ways than coping with anxiety. If you don’t use AI, you are probably disappointed that I would ever use a product that is killing our environment, stealing creative work, and making us dumber.

I want to reiterate that I am not proud of using AI, and I don’t support using it. But I have used it, and I understand the allures of it, and I believe it’s best to publicly admit that and share how I am trying to change rather than privately feeling shame. I want more of us to openly talk about how we use or don’t use AI, rather than only hearing from those entrenched in pro- and anti-AI camps. My bet is that a whole lot of educated people are floating in the middle, unconvinced by either direction while divisive rhetoric preaches to the choir.

Offline I know many people who live in the grey moral ground of using AI, allured by its speed, convenience, and comfort, while agreeing that it’s not good for our society. My partner and I often talk about this hypocrisy as it relates to politics—voters and constituents who support social welfare in theory but “not in their backyards.” AI is a political issue, and we are seeing hypocritical behavior with AI. I am a prime example.

To make real collective change happen, I believe we need to be better at speaking with this conflicted group of people without drowning them in shame. Shame makes people run towards AI, not away from it. We need to be able to have real conversations about AI at a human-to-human level to move towards moral action. I am grateful for the discussions I’ve personally had (thank you vrk, Jenamarie Boots, and Mel Mitchell-Jackson, to name a few folks) that finally convinced me to write this post and publicly commit to not using AI.

If you are using AI or know a loved one who does, I highly recommend watching the video How to (Anti) AI Better by Dr. Fatima. I agree with her research-backed take that AI literacy and harm reduction are the best tools for combating individual use of AI. She also points to collective actions we can take, like opposing new data centers (which are unfortunately often built in impoverished towns).

I also recommend Off the Grid’s AI podcast episodes, Cal Newport, and Ed Zitron for further AI commentary. I have not yet read Empire of AI by Karen Hao, but the book has been highly praised and recommended to me and I plan to read it soon.

Refusing the speed, efficiency, and dehumanization exacerbated by AI culture is possible. Our futures are not predetermined. The popularity of AI now does not mean it will inevitably take over the world. Whether or not these statements are true, I believe them in order to survive with hope. I believe them so I do not use AI.²

1. Not so fun fact: Korea ranks #1 in global AI slop consumption.

2. This ending was inspired by Ted Chiang’s short story “What’s Expected of Us”.