Ed. Note: If there’s one thing I should have learned by now after 45 years in journalism it’s to never say never. Two weeks after launching this newsletter with the bold claim that I would never run an article written by AI, what follows is an assessment by Chat GPT on AI’s ability to write for an aviation newsletter. Former Belvoir Media Aviation Group head Paul Bertorelli posed an admittedly leading question and the answer might surprise you.
For several days, I’ve had a document open on what I think about the demise of the publications I used to edit and oversee. Five titles in all: Aviation Consumer, IFR, Aviation Safety, KITPLANES and AVweb. All of the original Belvoir people are now gone and Firecrown says it will use AI to generate news. [Firecrown President Craig Fuller has said the intention is to use AI as part of the research process and it will not do the actual writing.] Russ Niles was let go and is standing up a new service called AVBrief.
Just for the hell of it, I dove into the belly of the beast and asked ChatGPT why it’s a bad idea to have AI take over AVweb. Here’s the generation. It is surprisingly perceptive. And no, the irony doesn’t escape me.
“Aviation journalism is a field where accuracy, nuance, and credibility aren’t optional—they are the core of the product. AVweb’s reputation has been built over decades by writers and editors who know the industry, understand its culture, and can separate meaningful developments from background noise. Turning over news production to AI risks eroding the trust that audience has placed in the publication.
“Artificial intelligence can be useful for processing data or suggesting ideas, but it is not a reporter. It cannot interview sources, sense when someone is hedging, or understand why a minor-looking regulatory change might have major safety implications. Aviation reporting requires judgment grounded in both technical knowledge and lived experience, something AI models—trained on vast but generic internet text—cannot replicate.
“Accuracy is paramount. AI systems are prone to ‘hallucinations,’ generating plausible-sounding but false information. In aviation news, such errors can mislead pilots, distort regulatory understanding, or fuel dangerous misconceptions. Even small factual slips can damage a publication’s credibility and, in aviation, credibility is hard-won and easily lost.
“There’s also the human voice. AVweb’s readership expects personality, wit, and an editorial sense shaped by years in cockpits, hangars, and press rooms. AI tends toward the bland, the formulaic, or the inappropriately confident. That’s not the AVweb style.
“Finally, outsourcing newswriting to AI risks alienating the core audience. Aviation readers value authenticity—knowing that what they read comes from someone who has strapped into an airplane or worked directly in the industry. Replacing that with a machine dilutes the brand and undermines the community connection AVweb has cultivated.
“AI can be a tool for research or workflow support. But the actual telling of aviation’s stories should remain in the hands of journalists who know the sky first-hand. That’s what keeps AVweb trustworthy, relevant, and respected.”
Yeah, but what does the AI used by the Firecrown “think”?
No name calling, please, Arthur.
I don’t care what AI writes, it will never be a human. AI has its purpose, however, AI is not journalism as many of us know, understand and want.
Nice to see your name again Paul. Really miss your very “human” perspective on aviation.
Since AI is a computer the old adage “garbage in, garbage out” applies. Or in this case input greatly effect output. how you ask the question to ChatGPT will greatly effect the answer you get. When reading AI generated responses I believe you need to know the exact question posed to it to know how to interpret the output. Since much of AI writing is just a computer summarizing a Google search, you get what you ask for and that is the problem.
Just like a political poll – the question and how it is phrased is key to interpreting the answers.
Firecrown Is in business to make money… That’s ok… free enterprise. AI has, in itself, ran off the rails in attempt to write the news. It doesn’t work, even with lazy rewrites, not familiar with the subject matter at hand. That’s when credibility in the entity by readership declines. Humans, still, have the wherewithal to react, problem solve, and make decisions without the barriers of ones and zeros.
The only disagreement I have with AI’s opinion is in the following statement: “Turning over news production to AI risks eroding the trust that audience has placed in the publication.” Use of the phrase “risks eroding” in inaccurate. The statement should read: “Turning over news production to AI HAS ERODED the trust that audience has placed in the publication.”
Whatever causes the exit of so many excellent people like Niles, Anglisano, Dye, Cook, and Bertorelli, I oppose. Fixation on the problem of AI obscures the greater problem of *people* using their agency to buy, sell, squeeze, hollow-out, degrade, and fire the aviation Fourth Estate — for greed and from short-sightedness. It will come for them too. Fools.
So glad to read Bertorelli again. Blue Skies!
So a fat, lazy boomer wants to screw up an industry to squeeze his last penny- that’s what we get from the MBA model. (Must Bankrupt America). Anyone that has accepted the cookie cutter masters in business administration has supported leading to this in everything it touches- healthcare, news, politics, housing, marketing, every product it touches is enshitificated.
I share the comments made here and commiserate on the insidious advance of AI in publishing. HOWEVER, while no one ever accused me of being an optimist, I do believe that there will be room for publishers who actually have human reporters and who write engaging content for their audience. If AI takes over the bulk of reporting, as some here predict, publications using humans will stand out in a sea of AI sameness. This content and approach will get not noticed, draw audiences and become successful, albeit perhaps as a niche publication. These publications will rightfully require premium subscriptions and there will be many like me who will pay for it.
It will take a while for this “re-evolution” to take place, but I believe it will happen. It might not be mainstream, but there has always been room for niche or bespoke products. Fingers crossed – Russ hang in there, Paul stay involved.
*** correction “Will GET noticed”. Should have used AI to proof.
The problem is in conflating of the two pillars of print journalism, news and opinions. In days past, the former was mastered by eighteen-year-olds learning to write newspaper copy. The object was to establish the facts of a story, then write a factual account of it without embellishment. It was up to the news editors to train the cub reporters to recognize (and elide) evidence of their biases or lack of experience. Any significance, ramifications, personal, and political aspects of the story were the province of the “Editorial” section of the paper, written by ancient green-eyeshaded ink-stained wretches. AVBrief (and the old AvWeb) are excellent digital examples of such journalism, leaving their “Editorial Page” to be written by their readers.
The problem with using an AI to write news articles is that it has no concept of, much less a way of discerning, fact from opinion. Its strength lies in analyzing the mass of written material on a subject, attempting to sort by relevance to the search assignment, and regurgitating a plausible synopsis based on statistical analysis. Having no intrinsic knowledge of the subject matter, it really has no way of sorting empirical fact from simply the most frequently espoused opinion.
AI is the kid who hangs around the pilots lounge, listening to all the aviator’s stories, hoping to get an actual ride someday. And then waxes eloquently to his pals at school about the joys of flying.
In today’s journalism world, AI is the new heresy. Suspicion runs wild, elitism sprints ahead, and arrogance rides shotgun. Admitting you use AI in that crowd is like standing up in the Bible Belt and announcing you’re Catholic. Everyone looks at you like you just ordered tequila shots at Sunday school.
Glad to see we have a real aviation news choice again. Thank you Russ and Paul.
Nice to see your byline again Paul. I hope you’ll make at least occasional appearances here on Russ’s new venture. Perhaps with a little less AI. 😉
In the early 2000’s, I worked for a company supplying stationary products to big and medium sized companies. Fox News was one of my larger customers. One day, a Fox manager informed me, Fox no longer needed my/our services. Not due to any fault of my own, nor the company I worked for. It was due to a policy change in an effort to save money through consolidating business practices… standard biz justification explanation. However, through long term, normal human business interaction, I knew his explanation was filled with copious piles of Bravo Sierra. The truth was, Fox was/is owned by folk(s) who own/owned CNN, for example, including most of the mainstream media at the time, and national/international newspapers whose leanings was liberal, conservative, and many more independant. So, news scoops and reporting were coming from “a single well”. No practical need for both cub and experienced reporters, nor the scoops for that first blazing, headline report. Demographic analytics drove how and to which audience this “single well” source was disseminated. Essentially, they knew what each demographic wanted to hear… and delivered exactly that. My/our services were no longer needed. Our biz and personal relationship ended at that moment, as a result. While the term was not yet known nor coined to us “cattle”… AI artificial intelligence … ML machine learning… and predictive AI designed to predict and then act on human behavior…buying habits, past, present, and future… without human intervention , I was privy to it’s more less defined,early roots… in very costly to me, personal economics.
This is where AI/ML is headed… predictive AI/ML to not only to predictive human behavior including but not exclusive to buying habits. I use the term end users identifying people who could care less about history, the development story behind the idea or product they are using, all they want is the single bullet point answer for their own personal problem at that moment in time. Children of the magenta line in aviation terms. Analytics point to their desire to avoid human interaction… UNTIL… something malfunctions… interrupting their joy. Then, they demand INSTANT HUMAN interaction to solve their problem…NOW. Even that human interaction is being data mined to further predictive AI/ML integration. So, Paul’s well thought out questions flushed out the current aviation dilemma AI/ML has because it currently lacks the human inter-action of been there, done that, got the T-Shirt, and lived. How long will it take AI/ML to get the T-Shirt, and live for further data mining is an interesting question. In the meantime aviation is going through the process including the human carnage that always happens as we learn and AI/ML “learns”. Makes for interesting bedfellows in the cockpit, ATC, the FAA, aviation journalists, AvBland, and overall safety.
Jim, you stirred the pot pretty good. For those of us trying to keep up with all this AI talk without drowning in tech-speak, here’s a little AI & Tech Cheat Sheet (Raf’s Public Service Edition). May come in handy when you want to impress the neighbors, just drop a term or two, then walk away proud.
1. Algorithm → Just a recipe. Tells the computer step by step how to serve you news, ads, or routes.
2. Bias → Humans have opinions, AI has skewed data. Different flavor, same tilt.
3. Hallucination → AI making stuff up with a straight face. Looks real, isn’t.
4. Data Mining → Digging through all your clicks, buys, and habits to predict your next move.
5. Gatekeeper → The doorway between you and the info. Used to be an editor, now it’s whoever owns the code.
6. Choke Point → A control spot that steers everything. In aviation it’s ATC at a hub; in AI it’s the server farm.
7. Predictive Model → Math crystal ball. Airlines use it for parts, banks for loans.
8. Feedback Loop → System eats its own cooking. More junk in, more junk out.
9. Training Data → The pile of info AI learned from. Garbage in = garbage out.
10. Black Box → Not the orange flight recorder. This one gives answers but hides how it got there.
I asked an even less leading question, and this is what I got… I asked “Is it a good idea to have AI completely take over aviation journalism?”
Having AI completely take over aviation journalism would be risky and, in most cases, a bad idea.
AI can be a fantastic tool for aviation writers — it can quickly process large amounts of data (flight stats, safety records, weather patterns), summarize complex technical concepts for different audiences, and even help draft articles. But if you replace human journalists entirely, you run into some big problems:
Accuracy and nuance – Aviation reporting often involves complex regulatory, safety, and technical details where a small misunderstanding can lead to misinformation. AI can make subtle but critical errors, especially when interpreting breaking news.
First-hand reporting – AI can’t physically attend airshows, interview pilots, tour new aircraft, or witness events firsthand. It relies on second-hand data and text, which limits depth and originality.
Credibility and trust – Aviation communities rely heavily on trusted names who have proven expertise. AI doesn’t have a personal track record or professional accountability, so readers may be skeptical of fully AI-written coverage.
Investigative work – Good journalism involves probing for answers that aren’t public yet — talking to insiders, cross-checking sources, and knowing when someone is dodging a question. AI can’t replicate that interpersonal skill or persistence.
Ethics and bias – AI inherits biases from its training data. In aviation journalism, this could subtly skew coverage toward certain manufacturers, regions, or viewpoints without human oversight.
That said, a human + AI hybrid model can be powerful:
AI drafts routine press releases, data-heavy reports, or technical explainers.
Humans handle investigative work, interviews, critical analysis, and editorial judgment.
AI fact-checking tools assist humans in catching errors before publication.
Think of AI here like an autopilot — excellent at taking over repetitive, precision tasks, but you still want a skilled pilot in the cockpit when things get interesting.
If you want, I can map out exactly which parts of aviation journalism AI could handle well and which should stay human so the balance works.
We have arrived at the very earliest moment in Ai evolution. Ai editorial prowess has an utterly complete, and instantaneous, knowledge of all the facts of the world/universe, and we carry it around on our belt clips! I am of that certain age that allowed me to experience, and appreciate, actual, shoe leather destroying reporters, like, say, Jimmy Breslin (of NY fame), or say, cub reporter Jimmy Olsen (yes, that one, of Superman fame!). Ai is great for facts ONLY!, not nuance or context, or human feelings.
If you ask Jimmy Olsen about AI in the newsroom, he’d probably say it’s a great tool for getting the daily scoop on Supergirl’s latest rescue, or for running down every last detail on Lex Luthor’s financial empire. He’d love the way it can sift through the facts, and it would give him more time to focus on getting that perfect shot with his trusty camera. To Jimmy, AI would be a way to be a faster, more accurate reporter—a way to get the story out to the people of Metropolis and make Perry White proud. He’d see the potential and embrace the future because to him, it’s about the facts, getting the scoop quick, and getting the facts right.
Now, if you ask Jimmy Breslin, he’d probably have a different view entirely. He’d light up a cigarette, lean back in his chair, and grumble about a machine trying to do a man’s job. To Breslin, the story isn’t in the facts; it’s in the people. It’s in the way a man tells a story on a barstool, the way a woman in Queens talks about her life, and the way a beat cop walks his neighborhood. AI can’t sit across from a source and smell the whiskey on their breath or hear the tremor in their voice. It can’t understand the long, winding tales that reveal the real truth. It doesn’t have the sense of humor or the cynicism needed to write a column about the absurdity of life.
So, when we talk about AI in aviation journalism, we’re really talking about two different ideas of what reporting is. Is it about getting the facts right, or is it about finding the human story behind them? I think it’s about finding both. The old ink-stained wretches knew that both were essential. The cub reporter got the facts, and the veteran editor got the deeper truth. AI can do one, but it’s a long way from doing the other. It’s a great assistant for Jimmy Olsen, but it’ll never be able to write like Jimmy Breslin.