How To Use AI without outsourcing your taste, thinking, or point of view.

AI can give you a clean, confident, perfectly structured draft in seconds.

Then somebody asks, “What do you actually believe about this?”

And your mind goes blank.

That is the strange problem with AI-generated work. The sentences can arrive before the thought does. Everything sounds finished, so you do not notice that an important part of the creative process never happened.

You never decided what was worth saying.

You never wrestled with the idea. You never chose an angle, rejected an easy answer, added a detail from your life, or noticed the contradiction that made the subject interesting in the first place.

Something was produced. But was it authored?

This guide will help you answer that question.

I am not anti-AI. I use AI for research, outlining, questioning, developing ideas, finding gaps, and editing. A good tool can become an extension of your mind. It can hold more information than your working memory, show you connections you missed, and give you feedback when you are too close to your own idea.

That idea has a philosophical name. In their 1998 paper The Extended Mind, Andy Clark and David Chalmers argued that, under the right conditions, tools outside the skull can become part of a cognitive process. A notebook can extend memory. A screen can make thought visible and rearrangeable. AI can extend reasoning too but only when it remains inside a loop you still direct.

But an extension of your mind should not quietly become a replacement for it.

That is the line this checklist helps you protect.

Authorship is not about typing every word yourself. It is about owning the meaningful decisions behind the work.

Use this guide before publishing an article, post, reel, newsletter, video script, presentation, or anything else you created with AI. It will show you where AI genuinely assisted you and where YOU may have disappeared from the work.

First, understand the four differences

These four distinctions are the foundation of AI authorship. Once you see them, a lot of “Is it okay to use AI?” debates become much easier to think about.

Prompting is not thinking

Prompting tells AI what to produce. Thinking decides what is worth producing.

“Write ten posts about discipline” is a prompt.

“I have noticed that my most consistent periods began when I made the daily requirement almost embarrassingly small. I want to explore whether discipline is sometimes a problem of emotional resistance, not ambition” is a thought.

The second one gives AI something to work with. It contains an observation, a tension, and a direction. The first one asks AI to supply all three.

A sophisticated prompt can still hide an empty idea. A simple prompt can develop a powerful thought. Prompt quality matters, but the quality of the thought underneath it matters more.

AI-generated is not AI-assisted

AI-generated work hands the answer to the machine. AI-assisted work keeps you responsible for shaping it.

Imagine AI gives you five possible hooks.

If you pick the first one because it sounds polished, AI generated the direction and you approved it.

If you reject four, combine two, rewrite the language, and choose the final hook because it creates the exact tension your audience feels, AI assisted a decision you still owned.

The difference is not whether AI touched the work. The difference is who made the decisions that gave the work its identity.

Fluency is not originality

Fluency makes an idea sound convincing. Originality gives it a point of view that came from you.

AI is very good at producing sentences that sound complete. That can fool you into believing the underlying idea is complete too.

It is not.

Originality does not mean inventing something from nothing. Every creator inherits language, formats, examples, techniques, and ideas from other people. Originality can live in what you select, what you connect, what you reject, how you transform it, and where you apply it.

The ingredients may be familiar. The relationship between them can still be yours.

That is also how you can learn from experts without becoming a copy: study the choices behind their work, keep what survives your judgment, and rebuild it inside your own context.

Output is not authorship

Output means something was produced. Authorship means you made and owned the decisions behind it.

An editor can improve your draft. A researcher can find evidence. A friend can challenge your argument. AI can help with all three. None of that automatically takes authorship away from you.

The real question is whether you stayed responsible for what the work says, why it says it, and how it reaches the reader.

If typing every word were the definition of authorship, transcription would be authorship. It is not.

The creative responsibility lives in the choices.

Find yourself on the authorship spectrum

AI use is not a switch with “pure human” on one side and “soulless machine” on the other. It is a spectrum of decision ownership.

AI replaced my thinking. ONE END OF THE SPECTRUM
AI produced → I approved
AI suggested → I selected
AI challenged → I reconsidered
AI helped me express what I meant
I still own the work. THE OTHER END OF THE SPECTRUM

You may move across this spectrum during the same project.

AI might generate an unimportant transition sentence while you own the core argument. It might help organize research while you decide what the research means. It might also write an entire section so smoothly that you accept claims you never examined.

The goal is not to perform every low-value task manually. The goal is to stay present wherever judgment is required.

Before you open AI, pass the intention check

The easiest way to preserve authorship is to arrive with something.

Not a complete thesis. Not a perfect outline. Just a raw piece of you: a question, observation, frustration, disagreement, personal experience, strange connection, or sentence you cannot stop thinking about.

Before opening AI, spend three minutes completing this:

Copy + paste: The intention check

WHAT I NOTICED:

WHAT I CURRENTLY THINK:

WHAT I AM STILL UNSURE ABOUT:

WHY THIS MATTERS TO ME OR MY AUDIENCE:

WHAT I WANT THE READER TO UNDERSTAND, FEEL, OR DO:

THE SPECIFIC DECISION I WANT AI TO HELP ME WITH:

Your answers can be ugly. In fact, they probably should be. This is thinking material, not publishable writing.

The point is to stop a blank prompt from becoming a blank mind with better grammar.

Your ideas need a home before they need an AI prompt

If your best thoughts are scattered across WhatsApp, Notes, screenshots, book highlights, and voice memos, start with The Scattered Ideas Rescue Kit.

The free kit gives you the four folders to create, copy-and-paste templates for each one, a 15-minute rescue plan, a 10-minute weekly reset, and a seven-day challenge for turning scattered inputs into ideas you can actually find and create from.

Get The Scattered Ideas Rescue Kit free →

The nine-part AI Authorship Checklist

Give yourself one point for every test your draft passes.

Do not treat the score as a moral grade. Use it as an editing map. A low score does not mean you cheated. It means the draft needs more of your judgment before it is ready to carry your name.

1. The origin test

Did this work begin with a question, observation, experience, problem, or argument that I genuinely care about?

Every authored piece needs a seed.

The seed does not have to be revolutionary. It could be a line you underlined, a problem you keep facing, something a friend said, a pattern you noticed at work, or an idea that annoyed you enough to investigate it.

What matters is that the project began because something caught your attention, not because AI said the topic was popular.

That seed usually appears when you pay attention long enough to notice what other people walk past. I wrote more about training that habit in See the World as an Artist.

Weak starting point:

Write a LinkedIn post about consistency.

Authored starting point:

Every time I built a long streak, I started with a target so small that skipping it felt ridiculous. I want to explain why lowering the minimum can create more consistency than increasing motivation.

The second starting point still needs development. But it already contains a human observation and a direction.

Repair the draft: Write one sentence beginning with:

“I wanted to create this because I noticed…”

If you cannot finish the sentence without asking AI, return to the topic before returning to the draft.

2. The thinking test

Did I form an initial thought before asking AI to form one for me?

Your first thought does not need to be correct. It needs to exist.

Thinking begins when you make a provisional claim, notice uncertainty, or choose a question worth following. AI can then challenge that thought, reveal gaps, show alternatives, or help you express it more clearly.

If you ask AI what to think before giving yourself a chance to think, its first answer becomes an anchor. You may spend the rest of the process decorating an idea you never chose.

This is why boredom and silence still matter in the AI era. If every unanswered question is immediately handed to a machine, you never stay with it long enough to discover what your own mind might do.

That unfinished gap is not dead time. It is fuel. The Dopaminergic Curiosity System explains why a question you notice and chase yourself can feel more rewarding—and become easier to own.

Repair the draft: Close the AI tab. Set a five-minute timer. Write continuously about the subject without looking anything up.

Then compare your notes with the generated draft. The differences are useful. They reveal where your curiosity naturally goes and where AI pulled the work toward a safer, more predictable answer.

3. The decision test

Can I name the important creative decisions I made?

Authorship leaves fingerprints.

You chose this argument instead of that one. You removed a clever line because it distracted from the point. You opened with a personal moment instead of a statistic. You moved a section because the reader needed an example before the explanation.

These decisions are the invisible architecture of the work.

If your entire process was “generate, skim, publish,” most of that architecture came from AI. Even if you changed a few words, the meaningful decisions may still belong to the machine.

Repair the draft: Name at least three decisions you made:

I CHOSE TO...

I REJECTED...

I CHANGED...

BECAUSE...

If you cannot name three, make three now. Change the order. Reject an angle. Choose a stronger example. Decide what the reader should remember.

Do not edit merely to prove that you touched the draft. Make decisions that improve what it means.

4. The point-of-view test

Does the draft reveal what I believe or does it merely explain the topic competently?

Information tells the reader what is known. A point of view tells the reader what you make of it.

Consider this sentence:

AI can help creators save time and increase productivity.

It may be true, but there is no author inside it. Almost anybody could say it.

Now consider:

The best use of AI is not producing more first drafts. It is helping you spend more time on better questions, sharper decisions, and the parts of the work that require taste.

Now there is a position. A reader can agree, disagree, test it, or remember it.

Your point of view does not need to be loud or controversial. It needs to show that a mind encountered the information and did something with it.

Repair the draft: Complete any two of these without AI:

• Most people misunderstand this because…
• The part I disagree with is…
• The advice works until…
• What my experience has shown me is…
• The overlooked connection is…
• If this is true, then we should…

Add the most honest answer to the draft. That sentence may become its center of gravity.

5. The specificity test

Does the draft contain details AI could not safely know or invent?

Generic writing talks about “creators,” “challenges,” “growth,” and “success.” Authored writing gives the reader something they can see.

It includes the exact sentence a customer used. The moment you noticed the problem. The book line that started the thought. The failed experiment. The screenshot, number, place, object, habit, or awkward detail that makes the idea real.

You can record a voice note, transcribe the whole thing, hand it to AI, and still receive generic khichdi. More input is not automatically better input. The useful material is the part that carries your observation, experience, or judgment.

AI can help you notice where an example is needed. It should not manufacture a personal story to fill the space.

Weak:

Many creators struggle to maintain a consistent writing habit.

Specific:

My “daily writing habit” failed every time it meant opening an empty document at 9 p.m. It became easier when I stopped trying to invent a post and started collecting one clear thought during the day.

Repair the draft: Add one piece of evidence only you can responsibly provide:

• A real moment you experienced
• A phrase your audience actually uses
• A result from your work
• A mistake you made
• A note, message, screenshot, or observation
• A source you genuinely read and understood

If the detail is personal, keep it true. If the claim is factual, verify it.

6. The taste test

Did I reject, remove, challenge, or reshape anything and do I know why?

Taste is not a list of adjectives in a prompt.

“Make it punchy, insightful, human, bold, and conversational” may change the surface of the draft. It does not tell AI what you consider good.

Taste becomes visible through selection and refusal.

You develop it by studying work you admire, understanding why it works, and noticing what weakens your own work. You use it when you say, “This sentence is technically cleaner, but it killed the energy,” or “This hook may get attention, but it promises something the piece does not deliver.”

AI can produce options. Taste decides which option deserves to survive.

Repair the draft: Mark three things:

• One line that feels alive and must be protected
• One line that sounds polished but says almost nothing
• One choice AI made that you would make differently

Then edit around the living part. Cut the noise around it. Do not sanitize the line that gave the draft its pulse aka, “Edit without Amputating”, that’s what author Henry Shukman like to call this process.

7. The voice test

Would I actually say this to another human being?

Your voice is not a magical tone setting. It is the accumulation of your decisions: the words you naturally reach for, the details you notice, the jokes you make, the amount of explanation you give, the metaphors you trust, and the places where you become blunt.

AI often smooths those edges because smooth language is easy to recognize as “good writing.” The result can be impressive and strangely anonymous.

Watch for sentences such as:

In today’s rapidly evolving digital landscape, creators must leverage the transformative power of artificial intelligence while preserving their authentic voice.

You might understand it. But would you say it?

A more human version could be:

AI can make your writing faster. It can also make everybody sound like they attended the same “boring” conference.

The second sentence is not better because it is casual. It is better because it contains a clear judgment in language somebody might actually use.

Repair the draft: Read it aloud.

Circle every phrase that makes your mouth feel unnatural. Replace it with the words you would use while explaining the idea to an intelligent friend.

Do not ask, “Does this sound human?” Ask, “Does this sound like me when I am trying to make this exact person understand this exact idea?”

8. The usefulness test

Does the reader know what to understand, question, decide, or do differently?

AI can explain a subject forever. Authorship requires deciding where the explanation is going.

A useful piece does not need a five-step framework. A poem can change what somebody notices. An essay can give language to a feeling. A story can complicate a belief. A tutorial can help somebody complete a task.

But the creator should know the intended change.

Without that decision, the draft tends to become a tour of the topic: definitions, benefits, challenges, tips, and a conclusion that politely repeats everything.

Repair the draft: Finish this sentence:

“After reading this, I want the reader to…”

Then inspect every section. Does it help create that change? If a paragraph is interesting but leads nowhere, move it to another note. It may belong to a different piece.

9. The ownership test

Can I explain and defend the work without reopening the AI conversation?

This is the final test because it contains all the others.

Could you explain the central idea in your own words? Could you tell someone why you chose this structure? Do you understand the sources? If a reader questions a claim, can you verify it, qualify it, or change your mind?

If your only defence is “AI gave me that,” the sentence is not ready to carry your name.

Ownership does not mean pretending to be certain. You can say, “I do not know,” “This is my current interpretation,” or “The evidence is mixed.” Those are authored decisions too. What matters is that you understand the level of certainty instead of inheriting AI’s confidence.

Repair the draft: Close every tab and explain the piece aloud in sixty seconds.

Include:

• The central claim
• Why you believe it
• The strongest example or evidence
• What AI helped you do
• What you personally decided

If you cannot explain one of those, return to the draft. Understanding should come before publishing.

Score your draft

Add one point for each test you passed:

• Origin
• Thinking
• Decision
• Point of view
• Specificity
• Taste
• Voice
• Usefulness
• Ownership

0–2: Mostly AI-generated

The machine supplied most of the idea, direction, and language. Do not waste time making tiny edits. Return to the intention check and decide what you actually want to say.

3–5: AI-shaped

You brought part of the idea, but AI made several decisions that define the work. Strengthen your point of view, add specific material, and reject more of the default structure.

6–7: AI-assisted

Your thinking is visible and AI helped you develop or express it. Repair the weakest tests, verify the claims, and make sure the final language still feels owned.

8–9: Clearly authored

Your judgment governs the work. AI may have accelerated research, options, organization, or editing, but the reader can still encounter a specific mind behind the piece.

Again, the score is not a purity test. An eight is not morally superior to a six. The question is whether the work is ready to represent you.

When AI has already swallowed the draft

Sometimes you realize too late that the draft sounds fine and contains almost nothing you chose.

Do not spend an hour replacing synonyms. Recover the thought first.

The point-of-view recovery exercise

1. Close the AI response. Do not keep its structure in front of you.
2. Write what you remember. Give yourself five minutes. No research and no editing.
3. Circle what still feels alive. Look for the sentence, question, disagreement, or example that carries energy.
4. Write what you agree and disagree with. Friction helps your point of view reappear.
5. Add one real detail. Use a moment, source, observation, mistake, or result you can stand behind.
6. Rebuild around that material. Use AI again only after you know what deserves to survive.

What survives after you close the tab is probably closer to your actual thought.

You may keep parts of the AI draft. The goal is not to destroy useful work. It is to restore a clear center of authorship.

Seven prompts that protect your authorship

Most prompt lists help AI produce more. These prompts help you remain responsible for what gets produced.

1. Use AI as an interviewer

I have a raw idea, but I do not want you to write the answer for me.

Ask me one question at a time to help me discover:
- what I actually believe,
- why I believe it,
- where the belief came from,
- what I may be overlooking,
- and what would make the idea useful to my reader.

Challenge vague answers. Do not draft the piece until I explicitly ask.

My raw idea:
[PASTE IT HERE]

2. Ask for directions, not a finished draft

Give me five genuinely different directions for developing this idea.

For each direction, show:
- the central question,
- the possible claim,
- the tension or contradiction,
- and what evidence or experience I would need.

Do not write the final piece. I want options I can judge.

My idea:
[PASTE IT HERE]

3. Turn AI into an intellectual opponent

Challenge this argument without rewriting it for me.

Identify:
- hidden assumptions,
- weak reasoning,
- missing perspectives,
- claims that need evidence,
- and the strongest reasonable disagreement.

Then ask me which criticism I want to address.

My argument:
[PASTE IT HERE]

4. Find the generic parts

Audit this draft for language or ideas that could appear unchanged in hundreds of articles.

Quote only the smallest relevant part, explain why it feels generic, and ask me for the specific observation, example, decision, or opinion needed to repair it.

Do not invent personal details.

Draft:
[PASTE IT HERE]

5. Expose the decisions AI made for you

List the meaningful decisions already embedded in this draft, including:
- the main claim,
- the audience,
- the structure,
- the examples,
- the tone,
- what was emphasized,
- and what was left out.

For each decision, ask whether I chose it deliberately or accepted it by default.

Draft:
[PASTE IT HERE]

6. Edit without amputating

Help me edit this draft without removing what makes it feel alive.

First identify:
- the lines with the most energy,
- the clearest original thought,
- the details that sound like me,
- and any messy-but-promising sections.

Protect those parts. Then suggest cuts or rewrites for repetition, filler, fake-smart language, and generic polish.

Draft:
[PASTE IT HERE]

7. Run the ownership check

Do not improve this draft yet.

Ask me questions that test whether I understand and can defend:
- the central claim,
- the reasoning,
- the evidence,
- the examples,
- the level of certainty,
- and the practical consequence.

Flag any sentence I seem unable to explain in my own words.

Draft:
[PASTE IT HERE]

The pattern behind these prompts is simple: AI asks, challenges, organizes, and reflects. You continue to decide.

Want a guided place to think this way?

If you want this kind of help without building a fresh prompt chain every time, I built CurioRabbit for exactly that.

CurioRabbit is a curiosity management and creation app. Start with one question, get a clear explanation, follow prerequisites and deep dives, compare mentor perspectives, reveal smart connections, and save the useful path as a journey.

It helps you understand an idea before you turn it into content without handing your point of view to a finished draft.

Explore your first idea free →

The AI-assisted creation workflow

Here is the complete process:

Notice → Think → Decide → Ask → Select → Rewrite → Verify → Publish

The arrows make the process easy to remember, but the process itself is a loop. In their 1981 paper A Cognitive Process Theory of Writing, Linda Flower and John R. Hayes described writing as a recursive movement among planning, translating, and reviewing not a clean march through fixed stages.

You may notice a better idea while rewriting, rethink the audience while verifying, or return to research after a claim breaks. Use this workflow as a map, not a conveyor belt.

1. Notice

Capture the raw material: a question, sentence, experience, problem, contradiction, or connection.

This part comes from living, reading, working, talking, paying attention, and occasionally standing in the shower when your brain decides to become a philosopher.

If that raw material currently lives across screenshots, highlights, voice memos, and old chats, my complete idea-capture and organization system shows how I turn scattered inputs into notes I can find, connect, and create from.

2. Think

Write what the idea currently means to you. You are allowed to be incomplete or wrong. Your job is to create something AI can respond to.

3. Decide

Choose the intended reader, the central question, and the change you want the piece to create.

4. Ask

Bring AI into a defined role. Ask it to question, research, challenge, organize, compare, or generate options.

“Help me” is vague. “Find the assumptions in this argument” gives the collaboration a job.

5. Select

Judge the response. Keep what serves the idea. Reject what flattens it. Combine useful parts. Ask better follow-up questions.

6. Rewrite

Move the work back through your own language. Add the examples, rhythm, judgment, and transitions the piece needs.

7. Verify

Check the sources, quotations, examples, names, numbers, and factual claims. AI confidence is not evidence.

8. Publish

Take responsibility for the final work. If your name is on it, the decisions are yours even when a tool helped you reach them.

If the work passes the checklist but fear still keeps it in your drafts, the next problem is no longer authorship. It is visibility. Stop Hiding Your ART is the push for that stage.

A quick before-and-after example

Here is a fluent AI-style opening:

In today’s rapidly evolving digital landscape, artificial intelligence is transforming the way creators generate ideas and produce content. By leveraging powerful AI tools, creators can streamline their workflows, enhance productivity, and unlock new levels of creativity while maintaining authenticity.

It is clean. It is also difficult to remember because it avoids making a meaningful decision.

Here is an authored version:

AI can give you five hooks before you have decided what you believe. That speed feels like progress. Sometimes, you are just overthinking faster.

What changed?

• The broad topic became a specific problem.
• The writer took a position on speed.
• The abstract promise of “authenticity” became a decision the creator must make.
• The language created an image instead of stacking business vocabulary.

AI could still help produce the second version. The sentence alone does not reveal the process. But the second version carries choices a creator can explain and defend.

That is the standard.

The 10-minute authorship audit

Open one AI-assisted draft you are considering publishing. Set a timer for ten minutes.

Minutes 0–2: Find the thought

Close the AI conversation and write the central idea in one sentence.

If you cannot, the draft has a topic but no center.

Minutes 2–4: Find the decisions

Name what you chose, rejected, and changed. Identify any major structural or argumentative decision you accepted from AI without examining it.

Minutes 4–6: Add evidence of a human life

Add one real observation, example, source, experience, phrase, result, or detail that makes the idea concrete.

Minutes 6–8: Remove generic polish

Read the opening, transitions, and conclusion aloud. Cut the sentences that sound impressive but carry no thought. Protect the lines that feel alive.

Minutes 8–10: Verify and own

Check the important claims. Then explain the piece aloud without looking at it.

Before you publish, confirm:

• I know what the piece is actually saying.
• I formed a view instead of borrowing one invisibly.
• I made the decisions that give the work its identity.
• I included specific material I can responsibly stand behind.
• I rejected or changed AI suggestions using my own taste.
• The language sounds natural in my mouth.
• I know what should change for the reader.
• I verified the claims that require verification.
• I can defend the final work without saying, “Because AI told me.”

Use AI without disappearing

AI is making execution faster. That makes clear thinking, taste, communication, and decision-making more valuable not less.

The advantage will not belong to the person who stubbornly refuses every tool. It will not automatically belong to the person with the longest prompt either.

It will belong to the person who knows what deserves to be made, uses the tool to expand what is possible, and remains responsible for the result.

You do not need to prove that you can create without AI.

You need to prove that you can use AI without disappearing from the work.

Open one unfinished draft. Score it. Fix the weakest test.

Make the work yours before you ask the world to give it their attention.

Until then,

Dewansh Jain

Frequently asked questions about writing with AI

How can I write with AI without losing my voice?

Form an initial thought before asking AI for help. Give the tool a specific role such as interviewer, researcher, critic, or editor then judge its suggestions, rewrite them in your own language, add details from your experience, and verify the final claims. Your voice survives when you continue making the decisions that give the piece its identity.

What is the difference between AI-generated and AI-assisted writing?

AI-generated writing asks the machine to supply the idea, angle, structure, examples, and language, while the person mainly approves the result. AI-assisted writing begins with human intent and keeps the person responsible for selecting, rejecting, changing, and defending the important decisions.

How do I make AI writing sound like me?

Do not solve the problem by asking AI to imitate a few surface quirks. Add what the machine cannot know by default: the observation that started the piece, what you believe, what you reject, a concrete detail from your life, the language you naturally use, and the consequence you want the reader to understand. Voice is the sound of repeated judgment, not merely a writing style.

Is it acceptable to use AI to help write an article?

Yes, if you remain responsible for the work. AI can help you research, question, organize, compare, edit, and generate options. You should still understand the central claim, verify factual material, make the meaningful creative decisions, and be able to defend the finished article without saying, “Because AI told me.”

What is the best way to use AI as a writing partner?

Ask AI to help you think instead of asking it to replace the thinking. Let it interview you, challenge an argument, expose assumptions, identify generic passages, find gaps, or present several directions you can judge. The strongest collaboration gives AI a bounded job while keeping authorship with you.