FIR #494: Is News’s Future Error-Riddled AI-Generated Podcasts, or “Information Stewards”?

Dec 29, 2025

Ask episode

Chapters

Transcript

Episode notes

In the long-form episode for December 2025, Neville and Shel explore the future of news from two perspectives, including The Washington Post‘s ill-advised launch of a personalized, AI-generated podcast that failed to meet the newsroom’s standards for accuracy, and the shift from journalists to “information stewards” as news sources. Also in this episode:

WPP founder Sir Martin Sorrell argued that PR is dead and advertising rules all.
Is AI about to empty Madison Avenue
Should communicators do anything about AI slop?
No, you can’t tell when something was written by AI
In Dan York’s tech report: Mastodon’s founder steps back, and new leadership takes over; the UN reaffirms a model of Internet governance that involves everyone: and Dan talks about what he’ll be watching in 2026, including decentralized social media, agentic AI, and Internet technologies.

Links from this episode:

Links from Dan York’s Tech Report:

The next monthly, long-form episode of FIR will drop on Monday, January 26.

We host a Communicators Zoom Chat most Thursdays at 1 p.m. ET. To obtain the credentials needed to participate, contact Shel or Neville directly, request them in our Facebook group, or email fircomments@gmail.com.

Special thanks to Jay Moonah for the opening and closing music.

You can find the stories from which Shel’s FIR content is selected at Shel’s Link Blog. You can catch up with both co-hosts on Neville’s blog and Shel’s blog.

Disclaimer: The opinions expressed in this podcast are Shel’s and Neville’s and do not reflect the views of their employers and/or clients.

Raw Transcript:

Neville Hobson Hi everyone, and welcome to the For Immediate Release long-form episode for December 2025. I’m Neville Hobson.

Shel Holtz And I’m Shel Holtz.

Neville Hobson And we have six great stories to discuss and share with you that we hope you’ll enjoy listening to during Twixtmas. What is that, you may ask? Well, Twixtmas is the informal name for the relaxed period between Christmas Day and New Year’s Eve, typically focusing on the 27th to the 30th of December. It’s a time for winding down, enjoying leftovers, watching TV, listening to podcasts, and simply existing without the usual hustle of holidays or work before the new year starts.

The name comes from blending Twixt, an old English word for “between,” and Christmas. It’s a modern term for a timeless lull in the calendar, often called the “festive gap.” That’s probably more information than you wanted, but now you know what it means. So, without further ado, let’s begin the Twixtmas episode with a recap of previous shows since the November long-form one.

Shel Holtz We’ll have to start using that over here.

Recent Episodes & Listener Comments

Neville Hobson That was FIR 489, published on the 17th of November. The story we led with in amplifying the long-form episode across social media was an explosion of “thought leadership slop,” where we riffed on a post by Robert Rose of the Content Marketing Institute. He identified idea inflation as a growing problem on multiple levels. Other stories in this 101-minute episode included quantum computing, vibe coding, “Is it OK to use an AI-generated photo in your LinkedIn profile?”, Dan York’s tech report, and more. And we have listener comments on this episode.

Shel Holtz We do, starting with Sherilyn Starkey up in Canada:

“I was just listening to the latest episode and you were commenting about a lack of female participation in podcasting. I thought I’d drop in a plug for my latest show, Stark Raving Social. I started it earlier this year and it delivers bite-sized episodes for marcomms pros. I do ‘how-to,’ ‘why you should,’ and ‘have you noticed’ type shows. I’m a hobbyist, so I publish when I have time and feel inspired, but it’s pretty regular. Last year I had a show where I interviewed 50 women over 50. And although the project’s complete, I still get about a thousand downloads monthly. I’ve been podcasting on and off since about 2007 and was—and still am—greatly inspired by FIR and your excellent work. Thank you.”

Thank you for that, Sherilyn, and hope to see you soon. Sherilyn’s terrific.

We have two comments on this episode from Darlene Wilson. She said:

“Enjoyed all of your content in this episode. Wanted to share that my role shifted from a marketing and comms managerial title to ‘Senior Manager, Corporate Brand and Communications’ a few years ago. It combines communication and brand leadership in one portfolio under which are marketing, sponsorship and events, promo, and change management. It’s a great role for a raging generalist. Moving brand and comms together—or brand under the comms umbrella—does signify part of a shift from end-deliverer of the message to a focus on reputation, trust, judgment, and the ability to oversee and connect what a company says and what it does. Given today’s environment, organizations do seem to want leaders, as Neville said, who bring judgment, sensitivity, and crisis literacy. That’s the comms person bringing broad and strategic thinking. Thank you both for your long-term commitment to this valuable profession.”

She added in another comment:

“The ‘every media revolution has slop’ analogy is directionally useful, but it can underweight what is genuinely discontinuous here: 1. Near zero marginal cost at massive scale, 2. Algorithmic distribution optimizing for engagement, and 3. Slop feeding back into training and ranking systems (i.e., model collapse plus search quality). If you treat it as just another cycle, you may miss that the mechanism is now self-reinforcing in ways Gutenberg-era pamphlets were not. The sources above—Google spam policies plus model collapse plus platform case studies—give you the evidence to make the distinction without turning the argument into moral panic.”

Neville Hobson Great comments. Thank you very much for that.

FIR 490 on the 1st of December: We unpacked some AI studies that claim to show what large language models actually read. But the sources shift month to month, and many citations aren’t reliable at all. We have a comment on this episode.

Shel Holtz From our friend, Niall Cook, who says:

“I don’t think anyone should be surprised that different studies report different results. It’s the same in many other research domains, but especially so here when the prompts, the models, the model parameters, and the methods will always produce differences—in the same way that no two users of the same generative AI system will get exactly the same response for the same question. We shouldn’t conflate visibility and citation reliability, though; two different things.”

Neville Hobson FIR 491 on the 8th of December shone a spotlight on big four consulting firm Deloitte, which created costly reports for two governments on opposite sides of the world, each containing fake resources generated by AI. Not only that, but a separate study published by the US Centers for Disease Control also included AI-hallucinated citations and the exact opposite conclusion from the real scientist’s research. We have a number of comments on this one.

Shel Holtz We have four, starting with Monique Zitnik:

“I’ve been nearly caught out with a source pointing to a website. After much digging, I discovered the website was AI-generated, and other websites had quoted this website. It was a myriad of AI-invented rubbish that sounded plausible.”

Mike Klein threw some praise your way, Neville. He said:

“It’s also a business model problem, as Neville pointed out in his excellent article for Strategic.”

That’s the magazine that Mike edits and you contributed to. He provided a link which we will add to the show notes; your article was titled Your Value is Not Your Timesheet.

Steve Lubetkin said:

“AI can be a useful tool, but humans need to review and confirm its output. The fact that they don’t or won’t is troubling.”

And Chris Lee wrote:

“You have both done some great episodes this year around AI. Very useful. Thanks. Keep them coming.”

Neville Hobson That was a great comment. Steve actually says it all: you’ve got to check up on all this stuff before you publish anything or rely on something. I see many more people now talking about it. You’ve got to verify everything all the time. You cannot trust it, whether it’s generated by AI or quoted by AI or linked to by the AI; you’ve got to verify all of that.

Shel Holtz Yeah, and I think we mentioned in one episode that I believe—and I think you do too—that there is likely to be a verification role that will be a new job classification. I’ve seen a little bit more about that since we made that assertion. There are actually companies that are hiring people to verify AI.

Neville Hobson That’s interesting, isn’t it?

In FIR 492 on the 15th of December, we looked at how the story of the untimely Omnicom layoffs in the US unfolded with one official investor-focused narrative and another on LinkedIn and Reddit. We observed that when people have platforms, the press release isn’t the whole story. We have one comment on this?

Shel Holtz Yes, from Roberto Capodici. Apologies if I pronounced that wrong. Roberto says:

“I think what’s really interesting here is how the whole situation highlights the tension between curated corporate narratives and the unpredictability of human experience playing out in public forums like LinkedIn.”

Neville Hobson In FIR 493 on the 22nd of December, we discussed how artificial early engagement can manufacture visibility that algorithms and media treat as significant. The tactics aren’t political; they’re portable and already familiar to communicators. It’s alarmingly easy to do.

And finally, we published an FIR interview on the 10th of December where we enjoyed a great discussion with Josh Bernhoff about his major survey of writers and AI. The deep divide between users and non-users, productivity gains, AI slop, trust, and the real story isn’t replacing people but resorting them. We have a comment or two, think, Shel?

Shel Holtz We have one. There are more on Josh’s repost of this. This one is from Susan Mangiero, PhD:

“I enjoyed your lively discussion about AI. In fact, I stopped the video and repeated several sections. I don’t think you addressed the use of AI for purposes of author marketing, unless I missed it:

What are your thoughts about using AI to help authors and their collaborating ghostwriters market their books?

Given Shel’s work in the area of employee communications, what are your thoughts about using AI for research? (Note: I do a lot of work with financial clients.) Josh, keep up the great work. I enjoy your blog. And the book survey was fascinating.”

Do you want to tackle these? I’m wrapping up work on a book right now. I have a proposal consultant helping me prepare the proposal, and I am thinking heavily about marketing these days. There’s no question that I will use AI as an aid to this in identifying targets to approach and testing language with different stakeholders. Every opportunity I have to use it to improve the marketing output, I will. I’m not going to outsource this to AI, but if AI can play devil’s advocate for me and help me brainstorm and ideate, I will take advantage of that all day long. What do you think?

Neville Hobson Absolutely, it is a natural tool to use. One of the biggest benefits of AI is its ability to literally be your right-hand person, your assistant who will work with you—not just respond to things you ask it, but challenge you on things. It’s the same as having a human being by your side, except this one doesn’t need to eat lunch.

It allows you to identify audiences, figure out what messaging is appropriate for which audience and when and where. It helps you concentrate on the next steps you’re to take.

Shel Holtz In terms of research for internal communication, I don’t see it as any different from research for external communication. It comes back down to the need to verify everything that you get.

I wrapped up a white paper for my company not too long ago on adaptive reuse of buildings. Since COVID, office occupancy has declined, and some large office buildings have defaulted on leases. The immediate thought is converting them to residences, but it’s harder than you think because of plumbing and natural light issues. The white paper explores other opportunities.

This is way outside my expertise, so I relied heavily on internal experts but also did a lot of research using Google’s Gemini Deep Research. I got a lot of great information, but some sources it found didn’t exist. I would have been humiliated if I had put out a white paper with that kind of information. I spent a lot of time verifying every source and every fact. It took less time than doing the research myself, but it was still time-consuming. As Steve Lubetkin noted, it’s disheartening that there are people who are not doing that.

Circle of Fellows Update

Shel Holtz I want to let everybody know about the most recent Circle of Fellows, which is now available for you to listen to or watch. It was a great conversation about the future of communication in 2026 and beyond. Zora Artis, Bonnie Caver, Adrian Cropley, and Mary Hills were the panelists.

The next Circle of Fellows is coming up on Thursday, January 22, at noon Eastern time. The topic is the impact of mentoring. We have a great panel: Amanda Hamilton-Attwell, Brent Carey, Andrea Greenhous, and Russell Grossman. You can tune in live or watch the replay on the FIR Podcast Network.

1. The Washington Post’s AI Podcast Debacle

Shel Holtz The core currency of a news organization isn’t its reporting; it’s trust. In mid-December, The Washington Post decided to trade that currency for a tech demo when it launched “Your Personal Podcast,” an AI-driven feature that generates audio summaries of the day’s news.

At its core, this doesn’t sound like a bad idea. Nicholas Negroponte suggested this in the 90s with the “Daily Me.” But at the Post, cracks appeared immediately. The AI mispronounced names, invented quotes, and editorialized. In one egregious example, AI announced a discussion on whether people with intellectual disabilities should be executed, stripping away the crucial context regarding a specific legal case.

According to internal documents obtained by Semafor, the product team knew exactly what they were releasing. During testing, between 68% and 84% of the AI-generated scripts failed to meet the newsroom’s own standards. In any other industry, a failure rate approaching 85% would trigger a recall, not a launch.

The Post is chasing a younger demographic that consumes audio, which is a valid goal. But serving them hallucinations doesn’t build a future audience; it alienates them. The Post needs to pull this tool, fix it, and apologize—not just for the errors, but for the decision to treat their subscribers as beta testers for a broken product.

Neville Hobson Extraordinary, truly. I was reading the NPR article you shared. It asks: “Will listeners embrace an AI news podcast?” The podcast is tailored to listeners based on what they’ve read in the Washington Post. That implies the likely listener is someone who spends a lot of time reading the Post, not a casual user.

It’s an intriguing step, but unfortunately, a misstep in terms of how they’ve dealt with it.

Shel Holtz Podcasting has become a staple for newspapers. The New York Times has The Daily and Hard Fork. Nothing is wrong with embracing podcasting. I just have a problem with the decision to launch it the way it was. The Washington Post is a storied institution—Katherine Graham, Ben Bradlee, Watergate, the Pentagon Papers. With this one decision, they have undermined that legacy.

Neville Hobson It symbolizes much of what is not right in the United States at the moment regarding freedom of speech and truth-telling. You mentioned Jeff Bezos owns the Post; where is the independence of journalists?

Shel Holtz We’re rapidly seeing this converted into state media, which is terrifying.

2. Martin Sorrell and the “Death of PR”

Neville Hobson Let’s talk about Martin Sorrell, the founder of WPP. On December 17th, in a debate on BBC Radio 4’s Today program, he declared the death of PR. Appearing with him was Sarah Waddington, the Chief Executive of the PRCA.

Sorrell made the blunt assertion that public relations is effectively dead and that the world has moved on to scale, reach, and speed—flooding the internet with content. Sarah Waddington pushed back firmly, anchoring PR in enduring purpose: helping organizations explain who they are and building trust.

The exchange was combustible, with Sorrell frequently talking over Waddington. Many felt Waddington was defending a way of thinking about communication that resists reduction to metrics alone.

Shel Holtz Every time Martin Sorrell opens his mouth, I roll my eyes. He once said WPP was more critical than human mortality. Advertising and public relations are not interchangeable. Advertising is about selling stuff; PR is about building relationships.

I always come back to the tuna boycott example. When StarKist addressed dolphin safety in their nets, PR agency Burson-Marsteller brought the parties to the table. The boycott organizers came out saying, “StarKist are the good guys.” Advertising could never have achieved that credibility.

Neville Hobson It feels like he was being provocative to generate headlines. But he seems to genuinely believe that scale, reach, and speed are what matter. If Sorrell thinks flooding the internet with detergent ads is the future, I think he’s crazy. I applaud Sarah Waddington for her calmness in the face of his bullying behavior.

Shel Holtz I challenge Sir Martin to find a client that will outsource their next existential crisis to WPP to handle with advertising. Let’s see how that goes.

3. The Future of Local News & Information Stewards

Shel Holtz The death of local news has been a consistent drumbeat. A new report from Northwestern University confirms news deserts have hit a record high. But a piece from the Nieman Journalism Lab argues the news hasn’t died; it just relocated to barbershops, church halls, and Facebook groups.

The Press Forward report suggests we look for “Information Stewards”—librarians, civic leaders, admins of neighborhood groups. If you’re a communicator, you can’t pitch a press release to a group chat, but you can provide clarity. Supply these stewards with fact sheets and FAQs. Trust has migrated from institutions to individuals.

Neville Hobson In the UK, local news is declining, though where I live in Somerset, there are three lively local papers. But generally, the commercial scale for local news is difficult. The idea of “Information Stewards” reminds me of the Epic 2015 flash video from years ago, which predicted a similar future.

Shel Holtz Local news is vital for accountability—school boards, zoning commissions. If no one reports on them, officials can do whatever they want. We need to reach these information stewards.

Dan York’s Tech Report:

Greetings, Shel and Neville, and all our listeners all around the world. Is Dan York coming at you from a snowy Shelburne, Vermont? And I want to begin this final episode of twenty twenty five, reflecting back on some of the topics and then upcoming changes with some of the things I’ve been talking about over these many episodes. A big one, of course, has been Mastodon and decentralized social media in general, and that had some big changes that have been happening in the past month.

Right around the time where we were recording the November show, there was a change at the the head of Mastodon. Now Mastodon is open source software, been around for ten years. That was created by a gentleman named Eugen Rochko, and he is the founder of this uh, has been based in Germany. And over time the organization evolved to be a German. Well, it tried to be a nonprofit, but then they’re a for profit entity. It’s they’re now in the scope of twenty twenty five. They have been looking to transfer to a, um, to a full non-profit, European based non-profit entity, most likely based in Belgium, according to the latest plans and all that, and they’re going through that process. But in the meantime, in late November twenty twenty five, Eugen announced that he will be stepping down as CEO and taking on a role as an advisor.

Now, this is critical because anybody who’s watched startups, whether they’re companies or whether they’re projects, knows that there’s a critical point when the founder needs to step away and let another management team come in and run the organization and grow. I have seen too many projects, including ones that I’ve led myself, where the founder, including myself, has stayed on too long and it just it dies at some point. Sometimes there are there are certainly cases where it has not, but there’s other times when it needs to move from the founder to others. So huge props to to Eugen and all the masks on folks for taking this step. And so there is a new leadership team. There’s a new executive director, a technical director, a community director, and there’s a team of employees and people who are continuing to evolve. Mastodon as one of the leading properties within the broader Activitypub based space that we call the Fediverse. So look for more to happen.

There’s a greater evolution going on over the scope of twenty twenty six. So cool things happening. I’ll note that this year, too many Mastodon servers just played on this whole wrapped thing, right? So you could get a wrap Staddon for twenty twenty five that wrapped up your most popular posts.

Some of the things you did, the most used hashtags, your archetype, all these different kinds of things. A little bit of fun just in the theme of all of the various different wrap things that are out there, but the fediverse will, I think, see a lot of activity and decentralised activity in general, because you’re seeing that through Mastodon and the other parts of the Fediverse. You’re seeing that with blue Sky and some tremendous work happening within the at, at protocol and some pieces that were there. Tim Chambers, somebody I come to really enjoy his writing over the time around open social items had a whole series of of predictions.

I’ll have the link for the show notes. He included some that were what he considered safe bets like blue Sky will cross sixty million registered users in twenty twenty six. He thinks he thinks the overall Activitypub fediverse outside of threads will cross fifteen million registered users, monthly active users, etc.. Uh, he’ll look at he had some ideas around threads. We’ll pass five hundred million. There’ll be continued federation. Anyway, if you’re looking for quick takes, it’s a good read. It’s some kind of interesting, fun stuff to think about and see where it will go around that.

Now, another story that I’ve been following this whole year has been internet governance kind of issues. And that culminated this month with a meeting at the United Nations called the World Summit on the Information Society. The twenty year review shortcut it has with this plus twenty. The good news coming out of all of that was that the governments of the world continued the path that we’re on, where everybody can be involved in some fashion in shaping the future of the internet, what was called in policy circles, the multi-stakeholder process.

But basically it means everybody has the potential to be involved in some way. It’s how the internet has worked since its origins. But there were some governments that wanted to put a different spin on it, where only governments would be involved and not businesses such as many of those of us listening to this, or universities or Or individual users. Anybody else like that?

So there were some good things that happened here. And something called the Internet Governance Forum, or IGF, has been made permanent rather than being renewed every ten years. It also had some other elements that recognize the the global network of national, regional and youth igfs that are happening all around the world. This is a venue, a way in which people, all of us listening, can be involved in internet governance. So it’s a great move, good step, lots of things. What am I looking at in twenty twenty six? You know, this whole episode is going to be about AI in different forms. I’ll be watching that too, specifically around the AI Agentic platform agents, the different pieces that were there and the different parts.

There’s a good article written by somebody over at the Open Future Foundation around why Wikimedia needs a seat at the Agentic AI Foundation, pointing out the work that happened in December that OpenAI and Anthropic and Block announced the creation of the Agentic AI Foundation, which also had Google, Microsoft, AWS, Bloomberg and Cloudflare joining into it. A lot of the commercial players all doing this. The point of this article was that it needs to have folks like Wikimedia involved and others. But in general, I personally will continue to be watching what’s happening at this agent level.

Agent to agent. Because that’s so much, I think, of what we’re going to be seeing as we increasingly look at AI driven tools and things. You know, that I’ll continue to be talking about decentralized social media, Mastodon, everything else. I’ll continue to be looking at, uh, internet and internet access.

You’ll hear me talk about low Earth orbit satellites, I’m sure, because we’re actually getting into a competitive situation where it’s more than just Starlink out there and also internet and information resilience. And so I want to leave you to a pointer, actually to a long read called landslide semicolon. A ghost story from Aaron Cassin, and I’ll have a link for the notes. But she writes a very long article piece around starting out about earthquakes, but getting into our information ecosystem and where we are, what’s out there, how it’s jumbled.

It’s worth reading and thinking about. Because really, the point is we need to think about how we have stories, how we work with things, how we have resilience in the information that we receive in some different forms. I encourage you to take a read about that. Think about it. Think what we will do in twenty twenty six. And with that, I wish you all a Happy New Year. I look forward to coming back at you in January. That’s all. You can find more of my audio writing at Dan York. Bye for now and back to you. Shel and Neville, Happy New Year.

5. AI Emptying Madison Avenue

Neville Hobson In a Wall Street Journal op-ed titled “AI is About to Empty Madison Avenue,” Rajiv Kohli of Columbia Business School argues that AI is quietly dismantling the agency model. Google, Meta, and Amazon are using AI to automate the advertising value chain.

While advertisers see efficiency, agencies see an existential threat. Madison Avenue isn’t being disrupted by better ideas, but by better systems. Kohli warns that unless things change, advertising may become a clear example of AI-driven creative hollowing out.

Shel Holtz I recently joined the advisory board for an AI certificate program at the University of San Francisco. The faculty stated they don’t believe AI will take jobs, which made me want to bang my head on the table. It already is.

Organizations need to strategize: what are the risks of outsourcing everything to AI? You can be efficient, but what do you lose? If you outsource everything, you’re going to see advertising overwhelmed with “slop.”

Neville Hobson The focus on speed and efficiency misses the important part: the people. We need to help educate leaders that AI should augment people, not replace them.

Shel Holtz In a capitalist society, leaders feel compelled to maximize ROI. If they can run a company with no employees and produce larger returns, they will. That’s why strategic analysis is vital to show where humans add value.

6. “Slop” is the Word of the Year

Shel Holtz Merriam-Webster has crowned “slop” as its Word of the Year for 2025. It defines it as digital content of low quality produced by AI. But a Scientific American article reminds us that every media revolution produces rubbish. The printing press produced libelous pamphlets; desktop publishing produced ransom-note newsletters.

The backlash isn’t a rejection of AI, but of low quality. To stand out in a sea of slop, your content needs to be exceptional.

Neville Hobson That Scientific American piece was great—calling Gutenberg the “ChatGPT of the 1450s.” It isn’t anti-AI; it’s about the sheer volume. If you automate production at scale, that’s flooding the internet, and much of it will be slop.

Shel Holtz You have to stay on top of the research. A study found people liked AI-generated ads more than human ones—until they were told it was AI. That shows an anti-AI bias, but also that the “human in the loop” matters for trust.

7. Detecting AI Writing

Neville Hobson There is a growing confidence that we can tell when something is written by AI. But in the Financial Times, Elaine Moore argues that most “AI tells”—like the use of dashes or words like “delve”—are just normal writing habits. Large Language Models sound human because they are trained on us.

However, Wikipedia has a field guide to spotting AI writing, looking for clusters of signals like vague abstractions. The debate is shifting from “Can we detect AI?” to “How much certainty do we really need?”

Shel Holtz If the writing meets our needs and is accurate, do I care if it was written by a human or a machine? Disclosure is going to be important for trust purposes.

Neville Hobson Trust is becoming ever more important. Finding a source you can trust—someone who verifies and doesn’t hoodwink you—is the key.

Shel Holtz I used Google Gemini to help find sources for my book, but I checked every single one. I saved time, but I kept the human in the loop.

Outro

Shel Holtz We hope you enjoy your Twixtmas. Please leave us a comment on LinkedIn, Facebook, Threads, or Blue Sky. You can email us at fircomments [at] gmail [dot] com or leave a voicemail on the FIR Podcast Network website.

Our next long-form episode will drop on Monday, January 26th. We will resume our short mid-week 30 for This episodes starting next week.

The post FIR #494: Is News’s Future Error-Riddled AI-Generated Podcasts, or “Information Stewards”? appeared first on FIR Podcast Network.