Tag Archive | AI Verification

The Journalist Is Still Responsible

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.

Artificial intelligence can help a journalist work faster.

It can organize information, summarize documents, compare sources, identify gaps in an argument, suggest questions, assist with editing and challenge assumptions the journalist may have overlooked.

But there is one thing it cannot do.

It cannot take responsibility for the journalism.

That responsibility remains human.

The journalist remains responsible for the work they produce.

Editors remain responsible for the decisions they make.

News organizations remain responsible for the systems they approve, the standards they set and the safeguards they require.

AI can become part of the process without becoming accountable for the result.

That distinction sounds simple.

In practice, it is not.

If an AI system produces a paragraph that eventually appears in an article and that paragraph contains an error, who is responsible?

If AI summarizes a document incorrectly and nobody catches it, who owns the mistake?

If a newsroom introduces an AI tool into its workflow without adequate safeguards, can responsibility simply be pushed down to the individual journalist?

These questions become more important as AI systems move deeper into journalism.

The answer cannot be that the technology is responsible.

Accountability still belongs to the people and organizations using it.

The Temptation of Speed

One of the most attractive things about artificial intelligence is speed.

Tasks that once required substantial manual work can sometimes be completed much faster with AI assistance.

Information can be organized quickly.

Long documents can be examined.

Ideas can be compared.

A first draft can take shape faster than it might have before.

For journalists working under deadlines, that can be incredibly useful.

But speed can create another danger.

The faster information appears, the easier it can become to accept it without examining it closely.

A polished answer can feel authoritative even when the underlying information is incomplete, misleading or wrong.

And an AI system does not have to invent a fact to cause problems.

It can omit crucial context.

It can emphasize the wrong detail.

It can flatten uncertainty.

It can steer attention toward one interpretation while neglecting another.

That is why human judgment becomes more important, not less.

AI does not eliminate the need for verification.

It increases the importance of knowing what needs to be verified and why.

Polished Does Not Mean Correct

One of the first lessons anyone working seriously with generative AI learns is that confident language does not guarantee accurate information.

An AI system can produce an answer that sounds professional, complete and perfectly reasonable.

It can also be wrong.

That creates a particular problem for journalism because credibility depends on the accuracy of what gets published.

A spelling mistake can be embarrassing.

A factual mistake can damage trust.

An unsupported accusation can create much more serious consequences.

The danger is not simply that AI makes mistakes.

Human beings make mistakes too.

The difference is that generative AI can produce those mistakes quickly and present them in language that already looks finished.

That means journalists cannot judge AI output by how polished it appears.

They have to judge it by the evidence behind it.

AI Output Is Not Evidence

This is one of the rules I have increasingly adopted in my own work: AI output is not evidence.

If an AI system tells me something happened, that does not make the statement true.

If it provides a date, statistic, historical claim or quotation, that information still has to be traced back to reliable evidence.

AI can help point me toward something worth investigating.

It can help organize what I find.

It can help compare multiple pieces of information.

But an AI-generated claim should never be treated as verified simply because the system produced it.

That distinction has changed the way I work.

Instead of asking only, “What is the answer?” I increasingly ask:

“What evidence supports the answer?”

Those are very different questions.

The first can produce information.

The second begins the verification process.

Human Oversight Has to Mean Something

There is a phrase that appears constantly in discussions about artificial intelligence:

Human in the loop.

It sounds reassuring.

But simply putting a human somewhere in the workflow does not automatically make the process responsible.

A journalist can still accept an AI-generated summary without checking it.

An editor can still approve a story too quickly.

A newsroom can still adopt an AI system without understanding its limitations.

Human oversight only works when the human is actually verifying, questioning and exercising judgment.

Otherwise, the human becomes little more than a rubber stamp.

The goal should not simply be to keep a person somewhere in the process.

The goal should be to make sure that person still has the time, authority and responsibility to challenge what the technology produces.

The Human Has to Push Back

Another lesson I have learned is that AI becomes more useful when I stop treating it as something that should agree with me.

If I present an argument and ask an AI system whether it is good, I may receive a helpful answer.

But I often learn more by asking it to attack the argument.

What am I missing?

What assumptions am I making?

What would a skeptical reader challenge?

What evidence would weaken my conclusion?

What alternative explanation should I consider?

That is where AI can become useful as an adversarial partner rather than simply a writing assistant.

But there is another layer to this.

The AI’s criticism also has to be challenged.

It can identify a weakness that does not actually exist.

It can suggest an alternative interpretation that is unsupported.

It can overcorrect.

So the journalist has to evaluate both the original argument and the AI’s response to it.

The human remains responsible for deciding what survives.

AI Is Not Just Another Spellcheck

Journalists have always used technology.

Search engines changed research.

Digital recorders changed interviews.

Transcription software changed note-taking.

Content-management systems changed publishing.

Spellcheck changed editing.

So why should generative AI be treated differently?

The difference is not that AI is the first technology journalists have ever used.

It is that generative AI can produce outputs that resemble reasoning, interpretation and writing.

It does not merely correct a misspelled word.

It can suggest an argument.

It can summarize a document.

It can rewrite a paragraph.

It can compare competing explanations.

It can recommend what information appears most important.

That gives it greater potential to influence editorial judgment.

And the closer technology moves toward those judgment-heavy parts of journalism, the more carefully journalists need to understand what the system is doing.

Judgment Cannot Be Automated Away

Journalism is filled with decisions that go beyond arranging facts.

Which facts matter most?

Which source is credible?

Which detail belongs near the top of the story?

Which claim requires additional verification?

When is there enough evidence to publish?

When should publication wait?

What context does the audience need?

Those decisions involve judgment.

AI may help inform that judgment, but it cannot remove human responsibility for the decision.

That becomes especially important when a story involves uncertainty.

Real reporting is rarely as clean as the completed article makes it appear.

Sources may disagree.

Documents may be incomplete.

Eyewitness accounts may conflict.

Statistics may require context.

People may have incentives to describe the same event differently.

A journalist has to work through those contradictions and decide what can responsibly be said.

AI can assist with that process.

It cannot own the decision.

Responsibility Is Bigger Than a Byline

I often think about this in very simple terms.

If my name appears on an article, I am responsible for what is in it.

But responsibility in journalism extends beyond the byline.

Editors, producers, researchers and others may shape journalism without their names appearing prominently on the final product.

Whether my name appears on the work or I contribute behind the scenes, I am still responsible for the decisions I make in producing it.

And organizations have responsibilities too.

A newsroom cannot introduce AI tools, encourage their use and then pretend every consequence belongs only to the reporter.

Management decisions matter.

Training matters.

Editorial standards matter.

The design of the workflow matters.

Responsibility is shared, but it does not disappear.

What That Means for This Book

Artificial intelligence will play a significant role in the creation of this book.

It will help me organize research.

It will challenge arguments.

It will help examine drafts.

It may suggest language that ultimately survives into the finished manuscript.

But it will not receive editorial authority over the work.

I will decide what stays.

I will decide what goes.

I will decide whether the argument reflects the evidence.

And I will be responsible for what is published.

If an AI system produces something inaccurate, misleading, unsupported or simply not good enough, I cannot defend the final result by saying:

“The AI told me.”

That is not an editorial defense.

A Different Kind of Partnership

None of this makes AI unimportant.

Quite the opposite.

Used carefully, it can become one of the most useful tools available to a modern journalist.

But the most productive relationship may not be one in which the journalist asks AI to do the thinking.

It may be one in which AI helps the journalist think harder.

Challenge this.

Compare these sources.

Find the weakness.

Show me another interpretation.

What am I overlooking?

Where does this argument fail?

Those kinds of questions turn AI from a shortcut into something closer to an intellectual sparring partner.

And that may ultimately be one of its most valuable roles in journalism.

Responsibility Remains Human

The tools journalists use have always changed.

The responsibilities underneath the profession have changed much less.

Verify the information.

Understand the evidence.

Provide context.

Be fair to the people involved.

Correct mistakes.

Do not publish something simply because it sounds convincing.

Artificial intelligence does not erase those responsibilities.

It makes them more visible.

The journalist may have a new assistant.

The newsroom may have faster tools.

The organization may be able to accomplish more in less time.

But when journalism reaches the public and someone asks, “Can I trust this?” the answer cannot be:

“The AI said it was right.”

The answer has to come from the work behind the story.

And responsibility for that work remains human.


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

I’m Taking the Nonfiction Writing Challenge Again-This Time With AI

By Cheval John | Vallano Media November 1st, 2026
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.

Twelve years ago, I participated in a nonfiction writing challenge.

Generative AI was not part of my writing process.

There was no ChatGPT sitting beside me, no AI assistant helping me organize research and no system I could ask to challenge an argument before I published it.

The process was more traditional: research, notes, writing, rewriting and editing.

Now it is 2026, and I am doing the challenge again.

This time, the blank page has company.

For the next 30 days, I will be working on a new book, Journalist and AI — Volume 1, while documenting what happens when artificial intelligence becomes part of the nonfiction writing process.

This is not an experiment in having AI write a book for me.

It is an experiment in what happens when a writer uses AI as a tool while remaining responsible for the research, reporting, verification, judgment, writing and final product.

That distinction matters.

Twelve Years Later

A lot has changed since 2014.

Writers already had search engines, digital publishing tools, social media and numerous technologies that had transformed how information was gathered and distributed.

But mainstream writers did not have the kind of easily accessible generative AI systems available today.

Now I can use AI to help organize research, examine competing ideas, identify weaknesses in an argument, compare information and assist with editing.

But that does not automatically make the writing better.

It does not automatically make the research accurate.

AI can accelerate good work.

It can also accelerate bad work.

It can help organize information more quickly, but it can also produce convincing errors, flatten a writer’s voice or reinforce a weak assumption if nobody challenges it.

That is part of what I want to explore over the next 30 days.

Why Do This Again?

I have already written books.

So this challenge is not about proving that I can finish one.

I am returning to it because the writing environment has changed so much that I want to experience the difference for myself.

Where does AI actually save time?

Where does it improve the work, and where does it create new problems?

Most importantly, can AI help a writer become more productive without sacrificing accuracy, judgment, transparency or individual voice?

Those questions are part of the reason Journalist and AI exists.

Instead of discussing them only in theory, I am going to put them to the test while writing the book itself.

This Is Not a Perfect Comparison

There is one thing I want to make clear from the beginning.

This is not a controlled experiment between 2014 and 2026.

AI is not the only thing that has changed since 2014.

I have changed too.

I have another 12 years of experience as a writer, publisher and researcher.

Publishing technology has changed.

The projects are different.

The way I organize information has changed.

So if I work faster in 2026 than I did in 2014, I cannot simply say AI caused the entire difference.

AI is an important variable, but it is not the only one.

That limitation is part of the experiment, and I do not intend to hide it.

What I can document is how AI affects the process I use today.

How AI Will Be Used

Artificial intelligence will be part of this project, but it will have defined roles.

I will use it to help organize ideas, analyze information, identify questions, compare sources, examine arguments and assist with editing.

I will also use it as an adversarial partner.

One of the most useful habits I have developed while working with AI is asking a different kind of question.

Instead of only asking, “Is this good?” I often ask:

“What is wrong with this?”

I use AI to red-team ideas and drafts — to search for weak arguments, missing context, unsupported assumptions and alternative interpretations.

But the AI’s criticism is not automatically correct either.

I still have to evaluate it.

The same principle applies to research.

AI can help me find patterns, organize information and work through material faster than I could manually.

But an AI-generated statement is not evidence.

If something matters to the book, it still has to be verified against reliable sources.

What AI Will Not Do

AI will not receive final editorial authority over this project.

I will

I decide what belongs in the book.

I decide whether an argument accurately reflects the evidence.

I decide whether the writing sounds like me.

And if an AI system produces something inaccurate, misleading, unsupported or simply not good enough, I am responsible for catching it.

A system capable of producing polished language quickly can also produce polished mistakes quickly.

Speed cannot become a substitute for judgment.

What Readers Are Going to See

For the next month, I want to make parts of this process visible.

There will be successes.

There will probably be mistakes.

There will be AI suggestions that improve the work and others that I reject.

There may be research that changes the direction of a chapter.

There may be sections that get rewritten completely because the first approach does not work.

And if AI wastes time, gives me weak information, sends me in the wrong direction or produces writing that I throw away, that belongs in the record too.

The goal is not to create the illusion that writing with AI is effortless.

The goal is to document what actually happens.

I will not be publishing every chapter of the book as I write it.

Readers will instead see selected parts of the process, observations from the experiment and what I learn along the way.

At the same time, Vallano Media will continue its regular Tuesday and Thursday publishing schedule, with additional articles published when news, reporting or other projects warrant it.

The book challenge will happen alongside that work rather than replace it.

My Rule for the Next 30 Days

The Write Nonfiction in November challenge does not require me to work on the project every single day.

That is a rule I am adding for myself.

From November 1 through November 30, I will work on Journalist and AI — Volume 1 every day.

Some days may produce a large amount of writing.

Other days may be spent revising, verifying information, reorganizing a chapter or questioning an argument that is not working.

The point is not to chase an arbitrary daily word count.

The point is to keep moving the manuscript forward.

Thirty Days

By November 30, I want more than a word count.

I want a complete first manuscript of Journalist and AI — Volume 1.

But I also want a record of how we got there.

How much did AI accelerate the process?

Where did human judgment remain indispensable?

Did greater speed allow more time for verification and deeper thinking, or did it simply encourage more production?

Did AI help preserve my voice, or did I sometimes have to push back against its tendency to make everything sound polished and generic?

And after 30 days, will I view the relationship between journalists and artificial intelligence differently than I do today?

I do not know all of those answers yet.

That is exactly why I am doing this.

In 2014, I participated in a nonfiction writing challenge using the tools available to me at the time.

Twelve years later, I am returning in a completely different technological environment.

The technology has changed.

The responsibility of the writer has not.

For the next 30 days, you are going to see the process.


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