Introduction to Co-Creative AI Artistry | Don't ask AI for answers, cultivate them through dialogue

Note: This article has been completely revised to coincide with the launch of the official Co-Creative AI Artistry website in July 2026.

Through repeated practice with AI, the concepts of Co-Creative AI Artistry have gradually become more organized. In this article, I will introduce them in a way that is easy to understand for newcomers, based on the current official definitions and framework.

Preface: Welcome to the World of Co-Creative AI Artistry

If I write a better prompt, I will get a better answer.
When people start using generative AI, many think this way.
Of course, writing clear instructions is important.

However, as I deepened my dialogue with AI, I realized one thing.
You don't have to create the perfect prompt from the start.
You can talk about half-formed thoughts, provide feedback on the answers you receive, and gradually cultivate your questions.

Instead of trying to extract the correct answer from AI in one go, humans and AI cultivate answers together through repeated dialogue.
I call the mindset and practical methods for doing this 'Co-Creative AI Artistry'.

What is Co-Creative AI Artistry?

The official definition of Co-Creative AI Artistry is as follows:

Co-Creative AI Artistry is not a collection of special prompts.
It is neither a way to leave everything to AI nor a way to blindly trust AI's answers.
Humans present half-formed thoughts to the AI, and use the AI's responses as material to think further.
Through this back-and-forth, questions and ideas that were initially vague are gradually given shape.
The important thing is not to stop the moment the AI provides an answer.

This is the fundamental principle that supports Co-Creative AI Artistry.

The Two Pillars of Co-Creative AI Artistry

Co-Creative AI Artistry has two pillars.

✅1. How to frame questions

The 'how to frame questions' here does not mean writing precise questions from the start.
Rather, it begins by placing thoughts that are not yet organized into a dialogue with the AI.

For example,

• 'I'm thinking about this, what do you think?'

• 'I can't explain this well myself, so please organize it for me.'

• "Are there any problems with the premises of this idea?"

• "What happens if we look at this from a different perspective?"

• "Please provide some counterarguments as well."

• "Are there any points I've overlooked?"

By returning such questions, your own thinking as a human will also evolve.

The important thing is not to just throw a question at the AI and stop, but to take your own reaction to the response and feed it back into the AI.

"I feel like that's not quite right."
"This part resonates with me."
"Why did you think that?"
"In that case, isn't there also this possibility?"

Through this kind of exchange, your questions will be updated.

You don't need to think that you must perfect a good question from the very beginning to get a good answer.
There is no need to think that way.

You can simply cultivate the question itself through dialogue with the AI.

✅2. Thread Management

Another pillar is thread management.
(*Here, 'thread' refers to a single chat conversation unit with the AI.)
In chat-based AI, the flow of the conversation up to that point is accumulated in a single thread.
Therefore, a thread is not just a conversation history.
It is a workspace for thinking where purposes, premises, tried ideas, discarded ideas, and questions that arose along the way are piled up.
I think of a thread as something like a conference room.
If you handle completely unrelated topics one after another in one conference room, the conversation will get tangled.
On the other hand, if you separate conference rooms by theme, it becomes easier to deepen the conversation while keeping the premises shared.
The basic rule is:

is.
However, 'one theme, one thread' does not mean you should continue using one thread forever.

When a topic branches off, move it to a different thread.
Even for the same theme, if the conversation becomes long and old ideas start to mix with current ones, create a handover document and move to a new thread.
In the Art of Co-Creative AI, I call the former 'branching' and the latter 'generational change'.

Try it out in 3 minutes first

You don't need any difficult preparation to try the Art of Co-Creative AI.
You can start with just these three things.

✅Step 1: Talk about what you are thinking about

If your purpose is decided, start from the purpose.
If your purpose is not yet decided, it is fine to start from a sense of discomfort, a sudden idea, or a vague feeling.

This level is sufficient.

✅ Step 2: Return the parts you got stuck on

Read the AI's response and return the parts that caught your attention as they are.

This response deepens the initial question a little.

✅ Step 3: Leave behind what you understood and what remains undecided

At the end of the dialogue, organize the following three points.

• What was understood this time

• What has not been decided yet

• What to think about next

If you keep these, you can start the next dialogue from where you left off.
A dialogue that didn't reach a conclusion is not a failure.
If the question has become clearer, that in itself is an achievement.

Example of a short dialogue

Human:

AI:

Human:

AI:

Human:

The initial question was, 'Is text written by AI my own text?'
But through dialogue,

• creative acts

• judgment

• responsibility

• dependency on AI

multiple points of discussion have emerged.
This is what it means for a question to grow.
It is not that the AI guessed the correct answer.
By the human receiving the AI's response, providing feedback on discrepancies, and updating the question, the issues that need to be considered have become clear.

It's okay to start with 'What do you think?'

There is a very simple question I often use.

It is only four words.

• 'What do you think of this idea?'

• 'What do you think of this structure?'

• 'I came up with this mechanism, what do you think?'

• 'What do you think of this way of phrasing it?'

'What do you think?' is not a strict instruction.
That is precisely why you can ask the AI for a free response.
Depending on the context, you will receive various responses such as evaluations, questions, improvement suggestions, or different perspectives.
If the response is not what you intended, you can simply add conditions at that point.

Instead of writing long prompts from the start, begin with short ones and focus them through dialogue.
This is also a fundamental way to use co-creative AI techniques.

Six Dialogue Styles for Cultivating Questions

In co-creative AI techniques, we organize the dialogue used to leverage AI as a 'partner to think with' into six styles.
These six dialogue styles do not categorize the entire field of co-creative AI techniques into six types. They are a sub-system that organizes the roles requested of AI and how to advance the dialogue in order to practice the first pillar of co-creative AI techniques: 'how to formulate questions'.
This is not a classification of six types of good prompt examples.

• What to bring to the AI

• What role to ask of the AI

• In which direction to advance the dialogue

are what these organize.
There is no hierarchy or fixed order to the six styles.
Choose them according to your current state or purpose, switch between them during the dialogue, or combine several of them.

☝️ Putting into words and shaping

✅ 1. Vague Verbalization Style (Reflective Partner)

This is a style where you speak your thoughts, feelings, or sense of discomfort—which you haven't yet organized—directly to the AI.

While having the AI return questions to you, you explore what is bothering you.
In response to the explanation provided by the AI,

by responding with these, the outlines of thoughts and feelings that had not yet been put into words begin to emerge.
This is a style of using AI as a 'mirror' to reflect your own thoughts.

✅ 2. Feedback Convergence Style (Refining Partner)

This is a style where you bring text, plans, or initial drafts you have created yourself to the AI and polish them while receiving feedback.

You receive improvements or alternative ideas from the AI, and decide

• what to revise

• what to keep

• What to cut

• Which direction to refine it in

are things for humans to decide.
"Convergence" does not simply mean summarizing briefly.
It means bringing scattered ideas closer to a form that fits the purpose.

☝️Expand and find meaning

✅3. Creative Partner Type

This is a style where you throw fragmented ideas or thoughts at the AI and expand your thinking while receiving its reactions.

Have the AI provide multiple possibilities, choose the ones that interest you, and then layer on more questions.
Human ideas and AI reactions combine, sometimes causing thoughts to leap in directions that weren't there at the start.

✅4. Narrative Partner Type

This is a style where you find the meaning and values within your own experiences and events by telling them to the AI.

By telling the events in order and having the AI return questions to you,

• why you still remember that event

• how your perception has changed between then and now

• how that experience influenced your way of thinking

become visible.
The "story" here does not just mean novels.
It is a style that clarifies the contours of your own thoughts by retelling your experiences. It can also be applied to novels, articles, and structure building.

☝️Question and verify the foundation

✅5. Analytical Partner Type

This is a style where you carefully confirm the prerequisites of a discussion or the definitions of words with the AI.

Check for conditions you are unconsciously setting, confusion between facts and assumptions, and leaps in logic.
Use AI as a partner to analyze the foundation of your thinking.

✅6. Experimental/Verification Style (Critical Partner)

This is a style where you have the AI provide counterarguments, opposing views, weaknesses, or alternative explanations for the hypotheses or claims you have made.

The goal is not to deny your own ideas.
By adding a critical perspective, you refine your hypotheses and claims to be more valid.
However, having the AI provide opposing views is a verification of your thinking.
Important facts should be verified separately through primary sources or real-world results.

You can switch styles along the way

In actual dialogue, you do not need to stick to just one style.
For example,

1. Use the Vague Verbalization style to put thoughts into words

2. Use the Idea Expansion style to broaden possibilities

3. Use the Premise Confirmation style to check for assumptions

4. Use the Experimental/Verification style to look for weaknesses or opposing views

5. Use the Feedback Convergence style to polish your plan

You can proceed in this way.
Combining multiple styles is considered a "Composite Style" in Co-Creative AI techniques.
A composite style does not mean using all six every time.

For example,

This question uses the Feedback Convergence style and the Premise Confirmation style simultaneously.

Choose based on your current state

| Current State | Suitable Style |
| ---------------------------------------- | ---------------------------------- |
| Not yet put into words | Vague Verbalization Style |
| Have an initial draft or text | Feedback Convergence Style |
| Have the seeds of an idea | Idea Expansion Style |
| Have an experience or event to talk about | Story Partner Style |
| Want to organize definitions or premises | Premise Confirmation Style |
| Have a hypothesis or claim to examine | Experimental/Verification Style |

You do not need to create sophisticated questions from the start.
Bring what you have on hand to the AI and reselect your dialogue style as needed.
The six dialogue styles are not meant to be used only after you have perfected a good question.

They are methods for cultivating questions through dialogue with AI.

How is this different from 'kabe-uchi' (bouncing ideas off someone)?

Having read this far,
some of you may have thought,
'Isn't that just bouncing ideas off an AI?'

That understanding is not wrong.
Bouncing ideas off an AI is a part of co-creative AI techniques.
Co-creative AI techniques do not deny the practice of bouncing ideas off an AI, nor do they claim to be a superior method.

It is an operational system for connecting ideas born from bouncing them off an AI to:

• Updating questions

• Thread management

• Verifying hypotheses and information

• Human judgment

• Organizing and handing over results

Sometimes a single session of bouncing ideas yields a good plan.
On the other hand, the more important the theme, the less likely it is to be finished in a single dialogue.
We cultivate answers little by little while thinking from different angles, researching evidence, and testing things in reality.

Just asking the AI again does not constitute verification

AI can confidently provide incorrect information.
Also, if the premises a human brings in initially are wrong, the entire dialogue with the AI may proceed based on those premises.
Therefore, in co-creative AI techniques, we emphasize not only cultivating ideas through dialogue but also verification.
There are three stages to verification.

✅1. Have the AI re-examine

Ask it to provide counterarguments, oversights, and alternative solutions.
This is the entry point for checking your thinking.

✅2. Verify with primary sources

For important facts such as laws, statistics, research, and official announcements, confirm them with materials independent of the AI's answers.

✅3. Try it out on a small scale in reality

For things that can be executed, try them on a small scale and observe the reactions and results in reality.
Asking the AI again is not external verification.
Self-checking by AI is the entry point for verification, and necessary facts should be confirmed through primary sources or real-world results.
And it is the human who decides how to handle the results of the verification.
It is not the case that the AI proposes, the AI evaluates, and the AI decides on adoption.
Humans who hold the purpose, values, and responsibility make the decisions to adopt, hold, or reject.

Carrying over the results of a thread to the next

It is not uncommon to simply leave a long dialogue as it is.
Even if good ideas are born, when you look back later, you don't know where they were written.
You also don't know what has been decided and what remains undecided.
If that happens, the valuable dialogue ends as a temporary conversation.

Therefore, at a good stopping point, have the AI create a handover document.
For example, ask it as follows:

If you paste this handover document at the beginning of a new thread, you can pass the thinking you have cultivated to the next dialogue.
In Co-Creative AI Techniques, we consider not just the answers themselves, but also the questions, the reasons for decisions, and the pending items as the results.
Distill the results of the dialogue and carry them over to the next thought process.
By doing that, your conversation with the AI becomes an intellectual asset.

From a tool for getting answers to an environment for thinking

Generative AI is a convenient tool that answers when you ask questions.
It is useful for writing, summarizing, researching, and brainstorming ideas.
However, its potential goes beyond just one-off Q&A.

Try talking about thoughts you don't even fully understand yourself yet.

React to the words that come back.

Convey any sense of discomfort.

Explore other possibilities.

Verify the grounds for the information.

In the end, the human decides.

And, carry over the process of thinking to the next step.

When this cycle is born, the AI becomes not just a tool for receiving answers, but an environment for thinking.
You don't need a perfect question from the start.
First, try talking about what you are thinking about right now.
And then, try asking a question back to the answer you received.

Co-Creative AI Techniques begin with that one sentence.

For those who want to know more about Co-Creative AI Techniques

The official definition, the two main pillars, practical methods, thread management, and handover documents for Co-Creative AI Techniques are systematically summarized on the official website.

[Co-Creative AI Techniques Official Website]

In this note, I will share not only the concepts and practical methods of Co-Creative AI Techniques, but also records of creations, hypotheses, thought experiments, and trial-and-error processes born from dialogue with AI.

[Co-Creative AI Techniques Magazine]

[YouTube Channel]

Read systematically in books

Co-Creative AI Techniques are also introduced in Kindle books.

• "ChatGPT is all you need to talk to"

• "Co-Creative AI Techniques"

• "The Real Way to Use ChatGPT"

• "Weaknesses of Co-Creative AI Techniques You Should Know"

Cultivate through dialogue, verify through testing, and decide as a human.
Don't just ask AI for answers; question, think, and cultivate answers together with AI.
That is the essence of Co-Creative AI Techniques.