I have spent most of my adult life around radio stations, and if there is one thing broadcasters have always been spectacular at doing, it is becoming terrified of technology approximately five minutes before we figure out how to make money with it. We worried about automation. We worried about voice tracking. We worried about digital editing. We worried about streaming. We worried about smartphones. We worried about social media. Now artificial intelligence has arrived, and once again parts of the industry have divided themselves into two camps: people acting like AI is going to save broadcasting and people proudly announcing they wouldn’t touch the stuff if you paid them.
I don’t belong to either camp.
My position is considerably simpler.
Use AI.
I do.
But if you’re going to use it professionally, learn what the hell you’re doing.
Because the biggest danger to broadcasting isn’t artificial intelligence.
It’s artificial intelligence combined with human stupidity.
And Lord knows radio has never suffered from a shortage of the second one.
😂
Can We Please Stop With The Purity Test
Every so often I hear somebody proudly announce that they don’t use AI in their content.
Maybe they don’t.
I have no reason to call them a liar.
But I’m increasingly curious what they mean by AI.
Because artificial intelligence and machine learning are becoming deeply embedded in the tools broadcasters use for audio production, transcription, mastering, research, content management and distribution.
The National Association of Broadcasters isn’t treating this as some fringe technology for computer geeks. In 2026, NAB is offering broadcasters training specifically covering AI and machine-learning tools used in sound mastering and audio production, including technologies associated with Adobe Audition, iZotope RX, Ozone and LANDR.
Think about that.
The industry’s largest trade association isn’t holding a seminar titled:
HOW TO HIDE FROM THE ROBOTS
😂
They’re teaching broadcasters how these tools work.
So I’m not terribly interested in conducting some ridiculous purity contest over who is AI free.
I’m interested in something much more important.
DO YOU KNOW WHAT YOUR TOOLS ARE DOING
That’s The Question That Can Save Your Ass
Because here’s where broadcasters can get themselves into trouble.
You ask an AI tool for information.
It gives you something beautifully written.
It has names.
Dates.
Quotes.
Statistics.
Maybe even sources.
It looks magnificent.
There’s just one tiny problem.
Some of it may be bullshit.
That’s not my opinion about responsible journalism. The Radio Television Digital News Association specifically warns broadcasters that AI systems can misplace facts, lose context or confuse consumers. RTDNA recommends that journalists independently verify AI-generated information, test tools before publication, guard against inadvertent plagiarism and establish human review before material reaches the audience.
Read that again.
HUMAN REVIEW.
Not:
ChatGPT said it.
Gemini said it.
Claude said it.
The transcription program said it.
The computer said it.
Who gives a damn what the computer said?
WHAT DID YOU VERIFY
Because your audience isn’t going to sue the chatbot.
Your GM isn’t going to call the chatbot into the office.
Your advertiser isn’t going to scream at the chatbot.
Your attorney isn’t going to ask the chatbot why something defamatory went across your transmitter.
They are going to ask you.
The Button Still Has Your Finger On It
This is the philosophy I want every broadcaster to understand.
AI can help create something.
AI can suggest something.
AI can research something.
AI can organize something.
AI can transcribe something.
AI can enhance something.
AI can generate something.
But the moment you put your call letters on it?
YOU OWN THAT DECISION.
That’s the dividing line.
Not whether a machine participated.
Whether a responsible human being remained in control.
Use It To Find The Door But Walk Through It Yourself
This is how I approach AI.
If technology points me toward a story, wonderful.
Now I investigate the story.
If it gives me historical context, terrific.
Now I verify the history.
If it tells me a broadcaster worked somewhere in 1997, I’m going to look for evidence that the broadcaster actually worked there in 1997.
If it gives me a quote, I want the original source.
If it gives me a statistic, I want to know where the number came from.
If it tells me somebody died…
DEAR GOD PLEASE VERIFY THAT ONE.
😂
AI should accelerate your curiosity.
It should never replace it.
And For The Love Of God Don’t Ask The Computer To Be Your Lawyer
This is another place broadcasters can get themselves into spectacular trouble.
AI can help you understand a legal issue.
It can help locate documents.
It can summarize complicated material.
It can help identify questions you should ask.
That’s valuable.
But when the issue involves defamation, copyright, employment law, political advertising, FCC compliance, contracts, intellectual property or somebody threatening litigation?
CALL THE DAMN LAWYER.
The United States Copyright Office has already spent years examining AI because the legal questions are substantial enough to require a multipart federal study covering digital replicas, copyrightability and the use of copyrighted works in AI training.
And one of its conclusions should matter enormously to anybody creating content.
The Copyright Office says AI can absolutely assist human creativity without automatically destroying copyright protection. But purely AI-generated expressive material doesn’t receive copyright protection merely because somebody supplied prompts. Copyright continues to depend upon sufficient human authorship.
There’s your clue.
KEEP THE HUMAN IN THE WORK.
Not merely because I think that’s creatively better.
Because the law still cares about human authorship.
Then We Get To Voices
Now we’re entering the part of AI where radio needs to be particularly careful.
Because audio is our business.
And voice cloning is getting frighteningly good.
Could I create an artificial voice that sounds like a real person?
Technologically?
Absolutely.
Does that mean I should?
HELL NO.
Not without permission.
Not without understanding exactly what rights are involved.
Not without knowing what that voice is going to say.
And certainly not because somebody in promotions thinks it would be hilarious to have an artificial Morgan Freeman voice read the car dealer spot.
😂
Don’t.
The U.S. Copyright Office concluded that unauthorized digital replicas present enough risk to recommend new federal protection against realistic but false digital recreations of individuals.
SAG-AFTRA’s approach is similarly instructive. Its AI protections for digital voice replicas emphasize specific informed consent, control and compensation, including agreements requiring performers to approve uses of their replicated voices.
That’s not anti-AI.
That’s responsible AI.
There is a massive difference.
Consent Is A Beautiful Word
Want to clone your own voice?
Have fun.
Want to create an entirely synthetic character?
Interesting.
Want a personality to license an AI version of their voice so the station can generate specifically authorized material?
Get the agreement right.
Want to take somebody else’s voice because you found three hours of their airchecks online?
NO.
Want to make a celebrity say something they never said?
NO.
Want to resurrect a dead radio personality without considering the estate, applicable law, contractual rights or basic human decency?
OH HELL NO.
Technology being capable of something does not create permission to do it.
That’s kindergarten ethics with a GPU.
And Don’t Think Regulators Aren’t Paying Attention
The federal government already has experience dealing with AI-generated voices.
The FCC determined that AI-generated voice cloning falls within existing restrictions on artificial or prerecorded voices in robocalls under the Telephone Consumer Protection Act.
The FTC has likewise warned that voice cloning creates risks ranging from consumer fraud to the unauthorized appropriation of professional performers’ voices.
Broadcasters should pay attention to the principle behind those actions even when a particular regulation doesn’t govern a specific piece of station content.
A SYNTHETIC VOICE CAN HAVE REAL CONSEQUENCES.
Treat it accordingly.
Here’s Another Rule
NEVER LET AI INVENT A HUMAN BEING’S WORDS.
If you’re writing a fictional commercial?
Fine.
If you’re making comedy and everybody understands the premise?
Different conversation.
But journalism?
News?
An interview?
A statement attributed to a real person?
No.
If John Smith didn’t say it, don’t put quotation marks around it.
If John Smith said something similar, don’t have AI polish it into something prettier and continue presenting it as a direct quote.
If you’re uncertain?
Go back to the recording.
That’s journalism.
The computer doesn’t get to improve history.
And Please Protect Confidential Information
This one isn’t sexy.
It’s also incredibly important.
Don’t blindly dump confidential documents, personnel information, private advertiser data, unreleased financial information, passwords, proprietary research, contracts or sensitive source material into some AI service because you want a quick summary.
Know the platform.
Know its privacy terms.
Know whether information is retained.
Know whether your organization has approved it.
Know what happens to what you upload.
RTDNA specifically includes privacy among the fundamental issues news organizations should evaluate before integrating AI into newsroom workflows.
The five minutes you save summarizing something aren’t worth spending the next five months explaining why confidential information went somewhere it shouldn’t have.
Transparency Is Not An Admission Of Guilt
This is another thing radio needs to get over.
Some broadcasters act like admitting AI was involved in anything is equivalent to announcing:
ATTENTION LISTENERS WE HAVE FIRED EVERY HUMAN IN THE BUILDING
😂
No.
Transparency can build trust.
RTDNA’s guidance says disclosure of AI use is generally preferable to nondisclosure and specifically encourages newsrooms to determine where disclosure belongs across websites, social media and on-air content.
That doesn’t mean I think we need a disclaimer every time software removes a breath from an audio file.
That would become ridiculous.
But material uses?
Synthetic voices?
Artificially created images presented in realistic contexts?
AI-generated video?
AI-created news elements?
Content where a reasonable person could misunderstand what is real?
TELL PEOPLE.
Trust isn’t created by pretending technology doesn’t exist.
Trust is created by refusing to deceive the audience about how you’re using it.
Radio Has Something More Valuable Than AI
This may sound strange coming from somebody telling the industry to embrace the technology.
But radio owns something no artificial-intelligence company can manufacture.
A RELATIONSHIP.
The listener trusts the morning personality because she’s been there every morning for eleven years.
They trust the meteorologist when the tornado warning hits.
They trust the farm broadcaster because he knows their county.
They trust the sports host because he’s lived through twenty seasons with the team.
They trust the local news anchor because they’ve heard that voice during elections, floods, school closings, tragedies and celebrations.
That trust took years to build.
AI can help you serve that relationship.
It can also help you destroy it in about 37 seconds.
One fake quote.
One synthetic clip presented as real.
One photograph that never happened.
One AI-generated obituary about somebody who is still very much alive and currently calling your front desk wondering why you’ve killed them.
That’s all it takes.
And That’s Why I’m Not Afraid Of AI
I’m afraid of lazy people using powerful tools.
There’s a difference.
A chainsaw is a fantastic invention.
I would also prefer the person holding it to have some idea what the hell they’re doing.
😂
Artificial intelligence is no different.
We should be training broadcasters.
NAB is already doing exactly that, offering dedicated AI education for radio professionals covering AI-driven audio tools and modern production workflows.
We should be developing station policies.
We should be talking about consent.
Copyright.
Verification.
Disclosure.
Privacy.
Voice rights.
Data security.
Editorial responsibility.
And we should be updating those policies constantly because the technology we’re discussing today may look primitive eighteen months from now.
Here Is My AI Policy In One Sentence
If you’re looking for some 47-page corporate manual, I’m going to disappoint you.
Mine starts here:
NEVER LET THE MACHINE MAKE A DECISION YOU ARE NOT PREPARED TO DEFEND AS A HUMAN BEING.
That’s it.
Everything else grows from there.
Did AI help write something?
Can you defend every fact?
Did AI create an image?
Can you explain what it represents?
Did AI reproduce somebody’s voice?
Did you have permission?
Did AI summarize an interview?
Did you check the recording?
Did AI recommend a programming decision?
Did a programmer still make the decision?
Did AI produce a commercial?
Does somebody know whether every claim in that commercial is accurate?
Did AI help create a newscast?
Did a journalist verify it?
Did AI tell you something sensational?
DID YOU CHECK
If the answer is no, don’t put it on the damn air.
This Is Where Management Better Pay Attention
AI cannot become the latest excuse to eliminate experience.
That’s where this gets dangerous.
The technology can make a great producer faster.
It can make a great writer more productive.
It can give a small-market broadcaster tools that would have cost thousands of dollars only a few years ago.
It can help a salesperson brainstorm.
Help production create.
Help newsrooms organize.
Help programmers analyze.
Help talent prepare.
Wonderful.
But if management’s entire AI strategy is:
HOW MANY PEOPLE CAN WE FIRE
then management has completely misunderstood the technology.
The goal should be to give good people superpowers.
Not remove every good person and leave the supercomputer running the station.
Because Somebody Still Has To Know What Good Sounds Like
This is the part nobody talks about enough.
AI can generate 100 sweepers.
Who knows which one is good?
AI can produce 50 morning-show topics.
Who knows which one fits Fresno?
AI can generate a thousand songs.
Who knows which one belongs between Sabrina Carpenter and Doja Cat at 4:17
AI can summarize the city council meeting.
Who understands which sentence actually matters to the people listening
AI can generate an entire commercial.
Who knows whether anybody will remember the client’s name
That’s the human.
That’s the programmer.
That’s the producer.
That’s the journalist.
That’s the personality.
That’s the creative director.
That’s the broadcaster.
THAT’S THE JOB.
So No I Am Not Putting My AI Tools Away
Not happening.
I am going to learn them.
I’m going to experiment.
I’m going to ask questions.
I’m going to discover where they make my work better.
I’m going to discover where they make my work worse.
I’m going to use them where they’re useful and tell them to go to hell where they’re not.
And when something matters?
I’m going to verify it.
Because that’s my name on the work.
On The Dial Perspective
The National Association of Broadcasters has described AI as simultaneously presenting opportunities for local broadcasting and serious challenges involving misinformation, deepfakes and the protection of trusted journalism.
That’s exactly right.
This isn’t a choice between innovation and integrity.
WE NEED BOTH.
Radio shouldn’t run from artificial intelligence.
Radio should become extraordinarily good at using it.
Learn the tools.
Train the staff.
Write policies.
Protect private information.
Verify the facts.
Respect copyright.
Get permission before touching somebody’s voice or likeness.
Keep humans responsible for editorial decisions.
Disclose material synthetic content when the audience could reasonably be misled.
And never confuse the computer made it with the computer verified it.
Those are not the same damn thing.
AI Is Powerful But Your Name Is Still On The Door
That’s where I’m going to leave this.
I’m not interested in shaming somebody for using artificial intelligence.
I’m interested in whether they’re using it responsibly.
I don’t care whether a writer started with a blank Microsoft Word document, a legal pad, a voice memo or an AI-assisted brainstorming session.
Is the finished work accurate?
Is it original?
Is it ethical?
Were people’s rights respected?
Was the audience deceived?
Can the person who published it defend it?
Those questions matter to me considerably more than whether a machine helped somewhere along the way.
Radio has survived every technological revolution thrown at it because broadcasters learned how to take technology and turn it into something human.
We can do that again.
Use AI.
Use the hell out of it.
Make better radio with it.
Just remember one thing when that incredibly confident machine hands you something that looks absolutely perfect:
THE COMPUTER DOES NOT HAVE A LICENSE TO PROTECT
THE COMPUTER DOES NOT HAVE A REPUTATION TO LOSE
THE COMPUTER DOES NOT HAVE AN AUDIENCE THAT TRUSTS IT
YOU DO.
Act accordingly.
— Steven Mills
Publisher
On The Dial
#OnTheDial #ArtificialIntelligence #AI #Radio #RadioIndustry #Broadcasting #RadioProgramming #RadioNews #BroadcastJournalism #LocalRadio #MediaEthics #GenerativeAI #FutureOfRadio #BroadcastInnovation