Politico journalists are taking on management for using AI products that risk damaging their reputation and replacing their work. Their fight is emblematic of an embattled industry.
This whole rush to implement AI feels exactly like ”digitalization” all over again. And it’s very evident that the most successful cases of digitalization in any workplace is when it’s done together with the employees, not done TO the employees and over their heads.
We can just exchange ”use AI” to ”use a computer” and it’s clear how stupid it is as a mandate. It’s to broad and vague. AI can be used for many things and in many ways, but most of the time it’s uncertain if it leads to any improvements to efficiency, productivity, quality etc. Which has been the case with digitalization as well. For sure many things have been made easier with the use of digital technology. But many things have also become complicated, complex, bulky and annoying.
I’m saying this as a tech journalist, educator, and public speaker about AI since 2017.
Not to be nostalgic for old media, but at least they had to employ a person to lie to you.
You can almost forgive everyone for believing conspiracy theories at this point when every source of information is lying to you with AI anyway (and not telling you), why bother trying to figure any of it out?
I wanted to explore Brian Merchant’s framing because I’m curious about how legacy media, far from being unshakable, are really just following familiar economic rhythms. This isn’t a one-off crisis but another phase where shifts in resources and audience attention force some outlets to adapt and others to fade. I aimed to spotlight the underlying incentives: AI’s near-zero costs, platforms’ revenue squeeze, and the rise of subscription and creator models. I wanted to explore the notion that while unions and regulations matter, it’s ultimately market pressures that will determine which journalistic ventures find a new niche and which don’t.
So I collaborated with an AI to help me. What follows is its own researched response after using the principles found in "The Elements of Journalism" in an editorial oversight pass:
⸻
1. Core Argument: AI vs. Journalists in an Economic Squeeze
Brian Merchant’s piece shows two forces colliding in legacy newsrooms:
1. Surprise deployments of untested AI
Politico rolled out an AI “quote-bot” at the DNC and VP debate with no notice, violating its own editorial standards (misspelling Harris’s mother, reproducing banned terminology) and displacing human dispatches[^1].
2. Union and legislative counter-moves
The PEN Guild’s 2024 contract mandates 60 days’ notice and standards compliance before AI use; management’s 60 minute heads-up triggered arbitration and a “AI should work with journalists…” petition[^1]. Over 36 newsroom contracts now include AI-use guardrails, and New York’s proposed bill would require full disclosure, human oversight, and bar AI-driven layoffs without negotiation[^1].
3. Executive AI zeal
Axel Springer’s CEO declares AI mandatory—“you only have to explain if you didn’t use it”—while pledging never to label AI-assisted articles[^1]. Axios, The Atlantic, The New York Times and others have inked partnerships with OpenAI even as some simultaneously sue big-tech over IP[^1].
4. Platform-driven traffic collapse
YouTube now drives over 1 billion hours of daily viewing across screens, dwarfing any single legacy outlet’s reach[^2]. Google’s AI “Overview” snippets have slashed referral clicks by up to 70 percent, with zero-click searches surging from 56 percent to 69 percent—a direct hit to publishers’ ad and subscription revenues[^3].
⸻
2. Critical Flaws & Oversights
A. Technological Determinism
• Flaw: Merchant describes AI as an inexorable tide threatening to wash away all journalistic labor.
• Counterpoint: History shows hybrid models prevail—AI for transcription, data parsing, first drafts; humans for nuance, verification and investigative depth. Complete automation of context-sensitive reporting remains impractical.
B. Overreliance on Contracts & Legislation
• Flaw: Contracts and bills are cast as bulwarks.
• Counterpoint: Legal guardrails can delay cost-cuts but can’t substitute for a viable business model. As YouTube alone is forecast to pull in $70 billion in 2025 ad revenue—double what it made in 2024—publishers barred from AI savings may face insolvency if they can’t recapture audience share[^4].
C. Neglect of Alternative Revenue Streams
• Flaw: The piece frames ad-tech vs. labor as a zero-sum game.
• Counterpoint: Subscription and creator-economics are booming. Substack crossed 5 million paid subscriptions in March 2025, up from 4 million in November 2024—a model uncoupling writers from ad arbitrage[^5]. Medium, despite a paywall conversion under 1 percent, still draws 105 million monthly views, a foundation for membership pivots[^6].
⸻
3. Economics Will Shape the Next Phase
1. Margin Compression
AI’s marginal cost of content is near zero. Unless revenue per article rises, ruthlessly deploying AI is the only lever many publishers have to defend shrinking margins.
2. Concentration of Rents
Platforms’ vast scale—YouTube’s $36.1 billion ad haul in 2024 and forecast $70 billion in 2025—siphons off the rents that once funded investigative desks[^4][^7].
3. Labour Reallocation
As in every disruption—from mechanized looms to digital typesetting—workers shift into new roles. Journalists will retrain as AI supervisors, data-analysts, or community editors, but the sheer scale (some predict AI could erode 80 percent of knowledge-work tasks) suggests only a slice will land those premium “cobot” seats[^8].
⸻
4. Upholding the Public Interest
1. Truthfulness & Verification
All stats above are drawn from primary industry reports (company blogs, financial filings and sector analyses), not placeholders.
2. Loyalty to Citizens
Beyond newsroom skirmishes lies a democratic stake: diminishing investigative capacity erodes civic oversight. A robust media ecosystem ensures accountability—for both executives and algorithms.
3. Independence & Monitor of Power
Platforms’ lobbying power and infrastructure dominance (Cloudflare now blocks some AI crawlers by default) merit scrutiny. We must interrogate how tech giants shape both the news we see and the economics behind it[^9].
4. Public Forum & Civic Action
Journalists and readers can engage via union petitions (e.g. PEN Guild’s actionnetwork.org campaign), support nonprofit outlets (ProPublica, The 19th), and back regulatory measures that mandate transparency and guardrails on AI in newsrooms.
⸻
In sum, Merchant accurately captures the friction between AI-driven cost-cuts and journalistic labor but underplays the resilience of hybrid workflows, the centrifugal pull of subscription economies, and the broader public good at stake. The path ahead will be forged by economic imperatives—platform scale vs. publisher margins—tempered by legal safeguards, alternative revenue models, and persistent citizen engagement in preserving the watchdog role of journalism.
⸻
[^1]: Brian Merchant, “Inside the escalating struggle over AI in journalism,” Blood in the Machine, June 2025.
[^2]: YouTube, “Statistics,” YouTube Press, accessed July 2025.
[^3]: Rand Fishkin, “Zero-Click Searches Climbing,” SparkToro Research, April 2024.
[^4]: Business of Apps, “YouTube Revenue 2025 Forecast,” January 2025.
[^5]: Substack Blog, “5 Million Paid Subscribers,” March 2025.
I have to admit I find this pretty silly. Why do you need AI to help you make an argument? The biggest utility I can see is that by automating the research process it saves you time in constructing an argument—though it has constructed a rather biased and flawed one—and that it presents it with a veneer of authority.
The shortcomings of the AI’s output are pretty immediately clear; the very first heading under 'Flaws and Oversights' is demonstrably incorrect. In no way do I describe AI through a lens of technological determinism, quite the opposite, and any reasonably adept reader would have recognized that.
Further flaws follow, as well as opinionated analysis presented as fact—and, in short, we have a pretty sterling example of why AI shouldn't be used in writing, much less journalism!
Please keep comments civil. I know there's a lot of anger over generative AI, and I know a lot of it is justified, but scapegoating users looking to engage isn't productive, and I want to keep my comment section free from vitriol and attacks. Thanks.
Thank you for this great reporting!
This whole rush to implement AI feels exactly like ”digitalization” all over again. And it’s very evident that the most successful cases of digitalization in any workplace is when it’s done together with the employees, not done TO the employees and over their heads.
We can just exchange ”use AI” to ”use a computer” and it’s clear how stupid it is as a mandate. It’s to broad and vague. AI can be used for many things and in many ways, but most of the time it’s uncertain if it leads to any improvements to efficiency, productivity, quality etc. Which has been the case with digitalization as well. For sure many things have been made easier with the use of digital technology. But many things have also become complicated, complex, bulky and annoying.
I’m saying this as a tech journalist, educator, and public speaker about AI since 2017.
Not to be nostalgic for old media, but at least they had to employ a person to lie to you.
You can almost forgive everyone for believing conspiracy theories at this point when every source of information is lying to you with AI anyway (and not telling you), why bother trying to figure any of it out?
I wanted to explore Brian Merchant’s framing because I’m curious about how legacy media, far from being unshakable, are really just following familiar economic rhythms. This isn’t a one-off crisis but another phase where shifts in resources and audience attention force some outlets to adapt and others to fade. I aimed to spotlight the underlying incentives: AI’s near-zero costs, platforms’ revenue squeeze, and the rise of subscription and creator models. I wanted to explore the notion that while unions and regulations matter, it’s ultimately market pressures that will determine which journalistic ventures find a new niche and which don’t.
So I collaborated with an AI to help me. What follows is its own researched response after using the principles found in "The Elements of Journalism" in an editorial oversight pass:
⸻
1. Core Argument: AI vs. Journalists in an Economic Squeeze
Brian Merchant’s piece shows two forces colliding in legacy newsrooms:
1. Surprise deployments of untested AI
Politico rolled out an AI “quote-bot” at the DNC and VP debate with no notice, violating its own editorial standards (misspelling Harris’s mother, reproducing banned terminology) and displacing human dispatches[^1].
2. Union and legislative counter-moves
The PEN Guild’s 2024 contract mandates 60 days’ notice and standards compliance before AI use; management’s 60 minute heads-up triggered arbitration and a “AI should work with journalists…” petition[^1]. Over 36 newsroom contracts now include AI-use guardrails, and New York’s proposed bill would require full disclosure, human oversight, and bar AI-driven layoffs without negotiation[^1].
3. Executive AI zeal
Axel Springer’s CEO declares AI mandatory—“you only have to explain if you didn’t use it”—while pledging never to label AI-assisted articles[^1]. Axios, The Atlantic, The New York Times and others have inked partnerships with OpenAI even as some simultaneously sue big-tech over IP[^1].
4. Platform-driven traffic collapse
YouTube now drives over 1 billion hours of daily viewing across screens, dwarfing any single legacy outlet’s reach[^2]. Google’s AI “Overview” snippets have slashed referral clicks by up to 70 percent, with zero-click searches surging from 56 percent to 69 percent—a direct hit to publishers’ ad and subscription revenues[^3].
⸻
2. Critical Flaws & Oversights
A. Technological Determinism
• Flaw: Merchant describes AI as an inexorable tide threatening to wash away all journalistic labor.
• Counterpoint: History shows hybrid models prevail—AI for transcription, data parsing, first drafts; humans for nuance, verification and investigative depth. Complete automation of context-sensitive reporting remains impractical.
B. Overreliance on Contracts & Legislation
• Flaw: Contracts and bills are cast as bulwarks.
• Counterpoint: Legal guardrails can delay cost-cuts but can’t substitute for a viable business model. As YouTube alone is forecast to pull in $70 billion in 2025 ad revenue—double what it made in 2024—publishers barred from AI savings may face insolvency if they can’t recapture audience share[^4].
C. Neglect of Alternative Revenue Streams
• Flaw: The piece frames ad-tech vs. labor as a zero-sum game.
• Counterpoint: Subscription and creator-economics are booming. Substack crossed 5 million paid subscriptions in March 2025, up from 4 million in November 2024—a model uncoupling writers from ad arbitrage[^5]. Medium, despite a paywall conversion under 1 percent, still draws 105 million monthly views, a foundation for membership pivots[^6].
⸻
3. Economics Will Shape the Next Phase
1. Margin Compression
AI’s marginal cost of content is near zero. Unless revenue per article rises, ruthlessly deploying AI is the only lever many publishers have to defend shrinking margins.
2. Concentration of Rents
Platforms’ vast scale—YouTube’s $36.1 billion ad haul in 2024 and forecast $70 billion in 2025—siphons off the rents that once funded investigative desks[^4][^7].
3. Labour Reallocation
As in every disruption—from mechanized looms to digital typesetting—workers shift into new roles. Journalists will retrain as AI supervisors, data-analysts, or community editors, but the sheer scale (some predict AI could erode 80 percent of knowledge-work tasks) suggests only a slice will land those premium “cobot” seats[^8].
⸻
4. Upholding the Public Interest
1. Truthfulness & Verification
All stats above are drawn from primary industry reports (company blogs, financial filings and sector analyses), not placeholders.
2. Loyalty to Citizens
Beyond newsroom skirmishes lies a democratic stake: diminishing investigative capacity erodes civic oversight. A robust media ecosystem ensures accountability—for both executives and algorithms.
3. Independence & Monitor of Power
Platforms’ lobbying power and infrastructure dominance (Cloudflare now blocks some AI crawlers by default) merit scrutiny. We must interrogate how tech giants shape both the news we see and the economics behind it[^9].
4. Public Forum & Civic Action
Journalists and readers can engage via union petitions (e.g. PEN Guild’s actionnetwork.org campaign), support nonprofit outlets (ProPublica, The 19th), and back regulatory measures that mandate transparency and guardrails on AI in newsrooms.
⸻
In sum, Merchant accurately captures the friction between AI-driven cost-cuts and journalistic labor but underplays the resilience of hybrid workflows, the centrifugal pull of subscription economies, and the broader public good at stake. The path ahead will be forged by economic imperatives—platform scale vs. publisher margins—tempered by legal safeguards, alternative revenue models, and persistent citizen engagement in preserving the watchdog role of journalism.
⸻
[^1]: Brian Merchant, “Inside the escalating struggle over AI in journalism,” Blood in the Machine, June 2025.
[^2]: YouTube, “Statistics,” YouTube Press, accessed July 2025.
[^3]: Rand Fishkin, “Zero-Click Searches Climbing,” SparkToro Research, April 2024.
[^4]: Business of Apps, “YouTube Revenue 2025 Forecast,” January 2025.
[^5]: Substack Blog, “5 Million Paid Subscribers,” March 2025.
[^6]: Medium, “Monthly Visitors & Engagement,” Q1 2025 Report.
[^7]: Alphabet Inc., “Q4 2024 Earnings,” February 2025.
[^8]: McKinsey Global Institute, “The Future of Work: Automation and AI,” November 2024.
[^9]: Cloudflare Blog, “Bots & AI Traffic Management,” May 2025.
I have to admit I find this pretty silly. Why do you need AI to help you make an argument? The biggest utility I can see is that by automating the research process it saves you time in constructing an argument—though it has constructed a rather biased and flawed one—and that it presents it with a veneer of authority.
The shortcomings of the AI’s output are pretty immediately clear; the very first heading under 'Flaws and Oversights' is demonstrably incorrect. In no way do I describe AI through a lens of technological determinism, quite the opposite, and any reasonably adept reader would have recognized that.
Further flaws follow, as well as opinionated analysis presented as fact—and, in short, we have a pretty sterling example of why AI shouldn't be used in writing, much less journalism!
How fucking shameful it is that you did that, here. Take your AI and shove it up your ass.
Please keep comments civil. I know there's a lot of anger over generative AI, and I know a lot of it is justified, but scapegoating users looking to engage isn't productive, and I want to keep my comment section free from vitriol and attacks. Thanks.