When you’re talkin’ to yourself
Or to your AI-self

featured image
A morning view while take a sip of McFlurry Biscoff at BEC.

First, I would like to quote following lines.

It might be related to loneliness, the absence of love, not sufficient health relationship.

But not that I want to share in this story. Conversation with AI is the thing and what I think about it.

Answer features

To have a brief and concised answer in 1–2 paragraphs only, this limitation request is included in the question.

The conversation

Me: Please answer in brief. I would like to start our conversation with following lines (GNR Estranged lines above). Is it healthy or benefecial to talk to yourself as a reflection or idea sharping process?

GPT-5.2: Those lines are from “Estranged” by Guns N’ Roses — and they capture that feeling of isolation beautifully. [..] (too long)

Me: Please re-answer only in a concise paragraph.

GPT-5.2: Yes—talking to yourself as a way to reflect or sharpen ideas is generally healthy and often beneficial. Conscious self-talk can clarify thinking, improve problem-solving, regulate emotions, and even boost performance by helping you rehearse decisions or conversations. It becomes concerning only if it turns into persistent negative self-criticism or feels uncontrollable and distressing; otherwise, reflective inner dialogue is a normal and useful cognitive tool.

It is interesting to find that you are not crazy when talk to yourself as long as it is conscious and you do in on purpose, and of course not uncontrollable.

Me: The talking is not loud with sound, but only as whisper or via writing? Or it is really real talking but without talking patner? Write on Medium

GPT-5.2: All of those forms are normal. Self-talk can be silent (inner dialogue in your mind), whispered or spoken out loud, or even done through writing (like journaling). They’re simply different ways of externalizing your thoughts. Speaking out loud can sometimes improve focus and clarity, while writing often deepens reflection. As long as you recognize the thoughts as your own and they’re not distressing or uncontrollable, any of these forms are healthy.

I am not comfortable with speaking out loud or whispering. I prefer journaling and inner dialogue in mind to the two previous forms. Perhaps there is still cultural perspective that muttering is attributed to someone with sort of mental problem.

Let us clarify this in the next dialogue.

Me: How to differ muttering of normal person which is talking to himself for thoughts externalization to person with mental problem? Any patterns or features to be aware of?

GPT-5.2: [..] (long and itemized phrases).

I think it forgets the previous limitation or characteristic request of the answer. Remind the AI about it.

Me: Limit the answer in 1-2 concise paragraphs.

GPT-5.2: Normal self-talk involves awareness and control: the person knows they’re speaking to themselves, the content is logical and connected to real situations, and they can stop when they choose. It may be silent, whispered, or spoken aloud, and it doesn’t cause significant distress or interfere with daily functioning.

GPT-5.2: It may signal a mental health concern if the person believes the voice is external, responds to unseen stimuli, expresses strongly delusional or paranoid content, cannot control or stop it, or experiences noticeable distress and life disruption. The main differences are insight, control, and impact on functioning.

Awareness and control are important. If you know that you are performing self-talk and you can control to start or stop it, then you are sufficient sane. But if not, please seek help.

Let’s move from self-talk to talk with AI, which is included in next dialogue.

Me: When person share his thoughts with AI until some point that it can resemble the person, or he write all of his thoughts in a file and then talk to it with RAG, can this talk be considered as self-talk?

GPT-5.2: Yes — in principle, that can still be considered a form of extended self-talk. If the AI is trained only on your own writings and is used as a mirror to reorganize, challenge, or refine your thoughts, it functions like an externalized cognitive tool (similar to journaling, but interactive). The key factor is that you understand the responses are generated reflections of your own material, not an independent consciousness.

GPT-5.2: It becomes psychologically different only if a person starts attributing independent agency, authority, or emotional dependence to the system beyond its tool-like role. When used intentionally and with awareness, it’s closer to structured self-reflection than to anything pathological.

I encounter a new term “extended self-talk”, which is a safe way in using AI as its tool-like role, not beyond it such as attributing emotional dependence or authority. The last uintentional act will be pathological to a person mental health.

Closing

I think the story is sufficient for demonstration of self-talk that can be advanced with the help of AI and it is still heatlhy.

Another aspect to notice that AI may forget to the answer characteristics that we requested previously. Remind it more than twice seems working. Further experiments can be conducted to assure it remember the request.

Actualy, the request is intended to have concised dialogues, which are easier to document or use further, e.g. as in this story.

Topics: Self Reflection; Talk To Ai; Self Talk.