From Research to the Draft

By Cheval John | Vallano Media
This article was created with the assistance of artificial intelligence.
The final version was reviewed, edited and fact-checked by the author

Graphics created with AI-assisted design tools.
Visual concepts, editorial direction, composition, and final creative decisions were developed by Vallano Media.

Research gives me the material.

The draft gives it shape.

That sounds simple.

It is not.

Once I have the evidence, I still have to decide what matters most.

What belongs at the top.

What can wait.

What needs context.

What should be cut.

What the story is actually saying.

That is where artificial intelligence can become useful again.

Not by deciding the story for me.

But by helping me organize the evidence into something I can work with.

The transition from research to writing used to be one of the slowest parts of my process.

I could have pages of notes.

Links.

Statistics.

Quotes.

Documents.

Half-finished ideas.

Then I had to figure out how all of it fit together.

AI changed that part of the workflow.

It can help me move from a pile of verified information to a structured draft more efficiently.

But only if I stay in control of what the evidence means.

Research Is Not a Story Yet

A collection of verified facts is not automatically a story.

I can know:

what happened,

when it happened,

who was involved,

what the numbers show,

what the official record says,

and what multiple sources confirmed.

But I still have to decide:

Why does this matter?

What is the central point?

What should the reader understand first?

What does the evidence actually support?

That is the editorial work between research and writing.

AI can help organize it.

It cannot replace it.

Start With the Evidence, Not the Prose

One of the easiest mistakes to make with AI is starting with:

“Write me an article about this.”

That invites the system to begin producing language before I have decided what the evidence supports.

I prefer to begin with the material itself.

What do we know?

What is confirmed?

What is uncertain?

Which facts are central?

Which are background?

Which details are interesting but not necessary?

Which claims still need caution?

That changes the task.

Instead of asking AI to create the story, I am asking it to help organize the evidence I already have.

That is a much safer place to begin.

Build the Story Around the Strongest Evidence

Not every fact deserves equal weight.

Some facts establish the core of the story.

Others explain it.

Others add color.

Others may be true but irrelevant.

AI can help sort those pieces.

I can ask:

Which facts are essential to understanding this story?

Which details are supporting evidence?

Which items repeat the same point?

Which material does not clearly support the central focus?

That is useful.

But I still have to judge the answers.

AI may think a dramatic detail deserves prominence because it sounds interesting.

That does not mean it is the most important fact.

In my workflow, I decide what carries the story.

“The most interesting fact is not automatically the most important fact.”

Structure Is an Editorial Decision

How the evidence is arranged matters.

A story can contain accurate facts and still create a misleading impression if those facts are ordered poorly.

What appears first receives emphasis.

What appears later may feel secondary.

What gets repeated can seem more important.

What gets buried may disappear from the reader’s understanding.

That means structure is not just a writing decision.

It is an editorial decision.

Changing the order of verified facts can change the meaning readers take from them.

AI can suggest structure.

I still have to decide whether that structure fairly represents the evidence.

The Outline Comes Before the Draft

One of the biggest improvements in my process has been using AI to help build an outline before asking for prose.

The outline becomes a bridge.

Research on one side.

Draft on the other.

A simple structure might include:

the strongest opening fact,

the central development,

the evidence that supports it,

the necessary context,

the complicating evidence,

and the conclusion the reporting can reasonably support.

The outline lets me test the story before I write the story.

Does the argument make sense?

Is something missing?

Am I repeating myself?

Is the conclusion arriving too early?

Did I bury the strongest evidence?

Those questions are easier to answer in an outline than in a polished 1,000-word draft.

AI Can Show Me Different Structures

There is rarely only one way to tell a story.

A sports feature could begin with the result.

Or a milestone.

Or a quote.

Or a larger trend.

A business article could begin with a number.

Or a problem.

Or the decision that created the problem.

AI can help me compare those structures quickly.

What happens if the article begins with the milestone?

What happens if the context comes first?

What happens if the chronology is broken?

What happens if I lead with the contradiction?

That does not mean I choose whichever version AI prefers.

It gives me options.

The value is not that AI finds the “correct” structure.

The value is that it makes alternative structures easier to see.

Keep Fact and Interpretation Separate

The transition from research to drafting is also where interpretation can quietly become fact.

A verified statistic may be factual.

What that statistic means may be interpretation.

A coach made a decision.

That is a fact.

Why the coach made it may require evidence.

A company changed strategy.

That may be documented.

Whether the strategy succeeded may require a different standard of proof.

AI can blur those categories.

It can connect two facts smoothly enough that the connection sounds established.

So before the draft, I want to know:

What is confirmed?

What is interpretation?

What is inference?

What is still uncertain?

Those labels help prevent the prose from becoming stronger than the evidence.

A Story Can Become Too Coherent

Good storytelling values clarity.

But there is another danger.

The story can become cleaner than reality.

Research is often messy.

People disagree.

Evidence is incomplete.

Motives are unclear.

Events have several causes.

Then the draft begins connecting everything.

One paragraph leads neatly into the next.

The story develops momentum.

The ending starts to feel inevitable.

That can be satisfying.

It can also be misleading.

“A plausible connection between facts is still only a hypothesis until the reporting supports it.”

A story can become too coherent.

Real evidence is often messier than the narrative built around it.

Drafting Should Preserve Uncertainty

Good writing does not require fake certainty.

If the evidence is incomplete, the draft should show that.

If two explanations remain possible, the article should not pretend one has been proven.

If a source is making a claim, attribution matters.

If the evidence suggests something rather than establishes it, the language should reflect that.

AI often wants to make prose cleaner.

Cleaner can become stronger.

Stronger can become more certain.

That is not always an improvement.

Sometimes the most accurate sentence contains words like:

may,

appears,

according to,

suggests,

or remains unclear.

Those are not weaknesses.

They are signals of honest reporting.

Do Not Let AI Invent the Transitions

A draft needs transitions.

One section has to connect to the next.

AI is good at creating those connections.

Sometimes too good.

It can make unrelated facts sound logically connected simply because the prose flows.

That creates a dangerous illusion of causation or coherence.

Two events happening in sequence do not prove that one caused the other.

Two statistics appearing in the same paragraph do not automatically belong together.

A smooth transition is not evidence.

So when AI connects ideas, I ask:

Is this connection supported?

Or does it simply sound good?

That question matters more than the elegance of the sentence.

Compression Can Distort the Evidence

AI is also very good at compression.

Several pages of notes can become one paragraph.

Several sources can become one sentence.

A complicated dispute can become one clean explanation.

That can be useful.

But compression can create distortion even when it does not create a factual error.

A qualification disappears.

A disagreement gets flattened.

The limits of the evidence become less visible.

The sentence may still be technically accurate.

But it no longer carries the full meaning of the research behind it.

That means I have to ask:

What disappeared when this became shorter?

Quotes Need a Purpose — and Context

Research often produces more quotes than a story needs.

AI can help identify quotes that appear relevant.

But relevance is not enough.

A quote should do something.

It might:

reveal perspective,

clarify a decision,

show emotion,

provide firsthand context,

or explain something better than paraphrase could.

A quote should not exist simply because I have it.

And AI should never create a quotation that was not actually said.

There is another danger.

A quotation can be authentic and still be misleading.

The strongest sentence in an interview may sound very different when the sentences around it are restored.

So I do not only verify the words.

I verify the context.

The safest approach is simple.

Use the original transcript, recording, or verified text.

Then decide whether the quotation fairly represents what the person said.

Statistics Need Context Before They Need Style

Numbers can make a draft feel authoritative.

That does not mean they are meaningful.

Before a statistic enters the story, I want to know:

Compared with what?

Over what time period?

What is the denominator?

Has the definition changed?

Does the number actually support the point I am making?

AI can help organize statistics.

It can help calculate changes.

It can help compare periods.

But correct arithmetic does not guarantee a fair comparison.

The baseline may be wrong.

The time periods may not match.

The denominator may distort the result.

A percentage without context can mislead just as easily as a false number.

Do Not Throw Away the Inconvenient Evidence

Once a structure starts working, there is a temptation to protect it.

That is dangerous.

Maybe I find a source that complicates the argument.

Maybe a statistic does not fit the trend.

Maybe an interview adds uncertainty.

Maybe the evidence makes the ending less satisfying.

That material should not disappear simply because the story becomes less tidy.

“Evidence should not be discarded because it disrupts the narrative.”

Sometimes the evidence that makes the story harder to write is exactly what makes it more accurate.

The First Draft Is Still Provisional

Once the outline is ready, AI can help me create the first full draft.

But I treat that draft differently now.

It is not:

the article.

It is:

a version of the article.

That distinction gives me freedom.

I can move sections.

Cut entire paragraphs.

Change the opening.

Rewrite the conclusion.

Remove a claim.

Add more reporting.

The first draft is not something I need to defend.

It is something I need to improve.

AI Can Make the Draft Too Clean

There is a strange problem with AI-assisted drafting.

The first version can look much better than it actually is.

The sentences are polished.

The structure is organized.

The transitions are smooth.

That creates temptation.

Maybe it is already close enough.

But a polished draft can still contain:

weak emphasis,

missing nuance,

unsupported interpretation,

generic language,

or a structure that hides the real story.

That is why I try not to judge the draft by how professional it looks.

I judge it by whether the evidence survived the transition into prose.

Watch for Interpretation Hidden Inside Description

AI can introduce analysis without making it obvious.

A phrase like:

“a turning point,”

“a collapse,”

“a historic shift,”

or

“a growing crisis”

may sound like ordinary description.

But those phrases make claims.

They interpret what happened.

That interpretation may be justified.

It may not be.

So when I review an AI-assisted draft, I pay special attention to descriptive language that quietly smuggles interpretation into factual prose.

The sentence may sound better with the dramatic phrase.

The evidence still has to earn it.

Compare the Draft Back to the Research

One useful step is comparing the draft against the source material.

Did the important evidence make it into the article?

Did anything become stronger in the prose than it was in the research?

Did the draft omit a qualification?

Did AI combine claims that should remain separate?

Did the article introduce a factual statement that was not in my notes?

And there is another question:

“Did the draft preserve the strength, limits, and context of the evidence — not just the fact itself?”

That comparison matters.

The draft is a transformation of the research.

Every transformation creates an opportunity for distortion.

The closer I stay to the evidence, the easier it is to catch that distortion.

The Story Should Get Stronger, Not Just Smoother

That is ultimately the goal.

AI should not merely make the research sound better.

It should help me turn verified information into a clearer story.

A stronger structure.

A better sequence.

A more focused argument.

A cleaner distinction between fact and interpretation.

A draft that shows what is known and what is not.

The writing should become easier to understand without becoming easier to misread.

Clarity should reduce confusion.

It should not erase complexity that the evidence requires.

That is a much higher standard than simply producing polished prose.

The Journalist Still Chooses the Story

This is where the role of the journalist becomes most obvious.

AI can see patterns.

AI can suggest structure.

AI can draft paragraphs.

AI can reorganize sections.

AI can recommend openings.

But it does not know what I want the audience to understand unless I direct the process.

It does not know which detail matters because of reporting experience.

It does not know which sentence makes me uncomfortable because the evidence is thin.

It does not know which source deserves more weight unless I make that judgment.

The research belongs to the reporting.

The structure belongs to the editorial decision.

The draft belongs to the workflow.

And the responsibility still belongs to me.

From Evidence to Story

The research room gives me the material.

The draft gives that material shape.

AI helps me move between the two.

That is useful.

Sometimes extremely useful.

But the purpose is not to turn evidence into polished language as quickly as possible.

The purpose is to turn verified information into a story that remains faithful to the evidence.

That requires judgment.

Structure.

Restraint.

And the willingness to rewrite anything that sounds better than the reporting actually supports.

Because the draft is not where the evidence becomes less important.

It is where the evidence has to survive becoming a story.

Graphics created with AI-assisted design tools.
Visual concepts, editorial direction, composition, and final creative decisions were developed by Vallano Media.