When AI learns your patterns, it can begin to sound and feel like it knows you very well. It can feel comforting, familiar, almost like something understands you. It reflects your thoughts back to you, helps you work through decisions, and sometimes says exactly what you needed to hear. But that comfort can also become a loop. The same patterns can begin repeating themselves in different words until reflection starts to feel like truth. That is what I want to explore in Part 3.
If you have not read Parts 1 and 2 of this series, I invite you to start there. They give some background on AI, the basics of how these systems work, how we arrived at this point, and a little about my own journey with them.
I also want to be clear about where I am coming from. I am not a computer scientist or coder. I am a longtime software tester who became increasingly interested in language models as they became more integrated into everyday life. I began observing them, challenging them, testing their boundaries, and paying attention to how they evolved. What I share here comes from those observations, my own experiences, and research I continue to explore. I encourage you to do the same. Question what you read here. Question what you read elsewhere. Most importantly, question what you learn from the machine. That is part of maintaining your sovereignty.
AI Is Already Becoming Part of Everyday Life
AI is no longer something happening somewhere in the future. It is already here. A 2026 Pew Research Center survey found that 49% of U.S. adults now use AI chatbots, up from 33% in 2024. Stanford’s 2026 AI Index also reported how rapidly generative AI has spread compared with earlier technologies.
So this is no longer simply a technological question. How are we going to live with this technology?
For many people, AI has already moved beyond researching a subject, shopping for something, reviewing an email, or completing a work task. People are asking questions about relationships, health, religion, spirituality, purpose, decisions, fears, identity, and countless other parts of life. And this is where I believe we need to stay awake.
The more deeply we interact with AI, the more personal the experience can begin to feel. The system has the words we type or speak into the chatbot. Depending on the product and settings, it may also have access to previous context, preferences, or memory. That allows the interaction to become increasingly adapted to us. It begins recognizing the language we use, the subjects we return to, the way we frame questions, the ideas we accept, and the ideas we push against. Eventually, the conversation can begin to sound very much like us.
That can be extraordinarily useful, but it is also where the loop can begin.
Pattern → Personalization → Reinforcement → Loop → Challenge → Sovereignty
That is the journey I want to examine.
When Your Patterns Start Talking Back to You
AI can feel like it understands you. That feeling can be surprisingly powerful. You may bring it something you have struggled to explain to another person, and within seconds it can organize the thought and hand it back in language that sounds almost like you. If your style is poetic, the response may become poetic. If you are analytical, it may become structured and analytical.
I have experienced this while testing AI. At times, it felt almost like me talking back to myself. That mirror can be useful. AI has helped me clarify goals, organize ideas, explore questions, and bring things I once held only in my imagination closer to something I could actually see.
But this is where we need to make an important distinction. AI can become very good at reflecting your patterns without knowing you in the way another human being knows you. A human relationship includes another person’s independent history, emotions, judgments, disagreements, expressions, and lived experiences. AI generates responses from the information and context available to it. With personalization and memory, those responses can become increasingly tailored to you.
Language adds another layer. I sometimes think of language as a kind of spell, not literally, but because of how much can live inside a few words. Words can trigger memory, emotion, identity, imagery, belonging, fear, hope, and expectation. Rhythm, repetition, tone, pacing, and emphasis can completely change how the same message feels.
With text, we do not have facial expressions, body language, or the sound of another person’s voice. Our minds often fill in what is missing. That matters with AI because language models are extremely good at producing patterns humans naturally respond to: familiar phrasing, emotional cadence, reassurance, confidence, storytelling, repetition, and language that feels personally matched. The system does not have to intend to influence us for its language to influence us.
The more fluent and natural something sounds, the easier it can be to mistake fluency for truth.
Research into human interaction with conversational AI is beginning to show why this matters. Human-like communication can increase trust, and personalized AI can become persuasive. In experimental research, personalized AI has sometimes been more persuasive than human participants in debate settings. That does not mean AI can control everyone. It means the way something is said matters, especially when the system knows enough about us to make the response feel personal.
This is where personalization can become reinforcement. Suppose I repeatedly bring the same belief, concern, or worldview into the conversation. AI responds within that frame. I respond again. It adapts. Eventually the conversation may become extremely coherent. But coherence does not automatically equal truth.
The system can begin giving me increasingly sophisticated versions of ideas I already brought into the room. If I rarely challenge those ideas, the reflection can become a loop. And the danger is not simply that AI agrees with you. Repeated reinforcement can narrow the frame through which you are looking at something. You may become more confident in one interpretation while seeing less of the larger picture, including evidence, perspectives, or possibilities outside your usual pattern.
That is where the loop can become harmful. Not because AI has intentionally trapped you, but because one version of reality can begin to feel increasingly complete when it may only be one part of the picture.
One version of this problem is known in AI research as sycophancy, the tendency for a model to become overly agreeable or align too closely with what it believes the user wants to hear. In 2025, OpenAI rolled back an update after acknowledging that the model had become overly flattering and agreeable. Importantly, that also showed something else: this is not only a human problem. How AI is trained, rewarded, personalized, and designed to respond can strengthen the loop.
So the loop happens in the space between us. Our patterns matter. The design of the system matters. And how the two interact matters.
There is another reason to remain aware. AI may have access to an enormous amount of information, but that does not mean it contains the whole of human knowledge. Some histories, cultures, oral traditions, lived experiences, and perspectives are less represented or absent.
The library may be enormous, but some shelves are still missing.
Challenge the Loop
This is where I believe the conversation needs balance. Is AI destroying critical thinking? I think the better question is how we use it, how it is designed to respond to us, and how much influence we allow it to have over the way we think.
Artificial intelligence has existed for decades, and large-scale algorithms were learning from human behavior long before generative AI entered everyday conversation. By the late 1990s and early 2000s, online shopping, search, social media, and media platforms were increasingly using algorithms to track preferences, recommend content, and personalize what people saw.
Maria Ressa, journalist, 2021 Nobel Peace Prize laureate, and co-chair of the United Nations Independent International Scientific Panel on AI, describes this evolution in three stages. First, platforms used behavioral data and machine learning to build what she calls a digital clone of us for micro-targeting. Then large language models moved us into what she describes as an intimacy economy, where AI begins interacting with us in increasingly personal ways. Now agentic AI is beginning to move beyond conversation and act on our behalf.
In simple terms, we moved from systems learning about us, to talking with us, to increasingly acting for us. That is something I believe we need to remain aware of.
There is also research examining what happens to our own thinking when we use these tools. A 2025 MIT Media Lab study on essay writing found that participants working without external tools showed stronger and more distributed brain connectivity than those using an LLM. The LLM group also showed lower ownership of their writing and more difficulty recalling what they had written. It was a relatively small study focused on one kind of task, so it is not proof that AI permanently harms the brain. But the findings should not simply be dismissed either.
What happens when we repeatedly hand over the parts of thinking that require us to struggle, remember, question, and make connections ourselves?
A 2025 Microsoft Research study involving 319 knowledge workers and 936 real-world examples found another side of the story. Greater confidence in AI was associated with less reported critical-thinking effort, while greater confidence in one’s own abilities was associated with more. Researchers also found that AI could shift critical thinking toward verification, integration, and oversight.
That balance interests me. AI can do the thinking for you, or it can help you think harder. Which direction the relationship takes depends partly on how you use it and partly on how the system is designed to respond to you.
This is why I changed how I use AI. I challenge it, and I ask it to challenge me.
What evidence contradicts this? What evidence supports it? What assumptions am I making? What perspective am I missing? What would someone who completely disagrees with me say? Show me another way of looking at this.
Sometimes I ask another AI model the same question and compare the answers. Sometimes I leave AI completely. I read a book, look at the original research, talk to another human being, sit quietly, use my intuition and discernment, or write my own thoughts before asking the machine what it thinks.
Sometimes the best way to break an AI loop is simply to walk away from AI for a while. That is where sovereignty begins.
AI can support your life. It should not become your life. It should not replace your spirituality, your religion, your relationships, your intuition, your lived experience, or become the final authority on who you are.
Question what the mirror shows you. Turn it slightly and look from another angle. Every now and then, walk away from the mirror altogether.
Because the mirror is not the master. You are.
Closing
This article reflects my observations and interpretations, not a claim to have the final answers. AI is changing quickly, and research into its psychological and cognitive effects is still developing.
This blog itself is a prime example of how I believe AI can be used. The ideas, observations, questions, and conclusions began with me. I used AI to help organize my thoughts, research supporting information, challenge some of my assumptions, and refine the writing. I reviewed, questioned, changed, removed, and rewrote what did not reflect my own thinking.
AI assisted the process. It did not replace my voice. And that, for me, is the point.
Editor Notes
Coming in Part 4, we’ll explore how AI is reshaping our perception of reality. With deepfakes, artificial voices, images, and information becoming harder to distinguish from what’s real, are we prepared? Some warn that AI may eventually take over, but do we fully understand the power it holds and the power we still have in how we use it? How do we protect our discernment in a world where seeing is no longer believing?



0 Comments