Bye Bye, AI: Why Publishers Should Think Twice Before Investing in AI

The Great AI Backlash of 2026 is in full swing. Since the debut of ChatGPT back in 2022, companies, including publishers, have tripped over their own feet in a savage rush to incorporate labor-saving AI platforms into their workflows. While the technology hasn’t yet reached the sophistication to trigger the much-heralded AI Job Apocalypse, the capital sucked up by investment in AI systems has triggered layoffs and reduced hiring in multiple industries and at all levels of employment.

Meanwhile, the social and environmental costs of data center operation and construction are the hottest topics in an already crowded political battleground, with Gallup reporting that 70% of Americans are opposed to the construction of new AI data centers in their local areas. The detrimental effects of AI on cognitive functioning are becoming clearer every day and the United States’ two largest school districts — Los Angeles and New York — have been forced to issue AI bans for students.

For publishers, the risks have been apparent from the beginning. Not only have AI summaries stolen valuable web traffic from online publishers, but AI models have simultaneously trained themselves using publishers’ copyrighted content. Add to that the open hostility that now greets generative AI content, and publishers would be crazy to risk their business models and reputations on the adoption of new AI agents.

And yet, publishers continue to outsource critical editorial functions to AI platforms. Just this week, the AI workflow brand Impelsys announced major investments in its mon’k and KAI platforms, which it advertises as “aimed at helping publishers move from isolated AI use cases to AI-native operations.” In non-buzzspeak terminology, that translates to AI-generated story ideas, writing, editing and production — what we have traditionally referred to as “publishing.”

For publishers who have already invested in generative platforms, it’s time to recognize these systems for what they are: expensive liabilities. The temptations for staff to use generative AI for purposes beyond a seemingly-innocent first draft or an interim proofread are all too real. And so are the risks:

1. Authenticity. Half of consumers already prefer to avoid generative AI content, with Gen Z being especially critical of the results. And AI-generated content has never been easier to detect. According to Derek Newton of Verify My Writing (which offers writers badges of authenticity to prove themselves the true authors of their own work): “As the AI has gotten better, the AI detection has also gotten much, much better. Now, a handful of technology platforms are really, really accurate at detecting AI text — accurate to the 1/10,000 or in some cases, 1/25,000 success rate.”  

2. Talent Acquisition. Many large- and medium-scale publishers have outsourced their human resource functions to AI-supported Applicant Tracking Systems, like Workday or Greenhouse. Besides accusations of inherent bias in these systems, they’ve effectively rendered the entire résumé-based hiring model obsolete. Algorithms within these platforms scan for keywords which have been seeded within applicants’ résumés by ChatGPT and the like, nullifying their own value. The best way to acquire talent remains the stubbornly old-fashioned method of building creator networks over time.

3. Expertise. Generative AI investors confidently aver that the responsible uses of these products occur at the very top and very bottom of workflows: mapping out strategy (at the top) and mindless, repetitive tasks (at the bottom). Apparently, the belief is held that enough honest human-powered work remains in the meaty middle between those extremes to maintain both quality and an adequate workforce. While that may be true, it elides an important constituent of human-powered work: experience. Generative AI is incapable of original thought; its recommendations are predicated on patterns of existing ideas. Every publication is, in its own way, a niche publication. The strategy that works for Rolling Stone cannot possibly benefit Scientific American, nor should it. And how will employees ever reach that safe (as advertised) middle ground without a foundation in the fundamentals of magazine publishing, be it copy fitting, or fact-checking, or color corrections?

Almost every rant against AI — or any other disruptive technology — seems to include some variation of the phrase, “I’m not anti-AI, but…” I’ll come right out and say it: I’m anti-AI. I haven’t come to this attitude without experience. I’ve used AI in my creative work and I regret every instance of it. I’ve even trained AI models and been left singularly unimpressed with the outcomes. And for those suspicious of my repeated use of em dashes in this post, all I can say is that I refuse to alter my writing style to contrast the preferences of a mindless algorithm that’s speedily poisoning our public discourse. For all I know, AI learned it from me.

The AI backlash is real. For publishers, adopting standards and practices around AI usage is the very lowest bar to clear. It’s time to stop feeding the monster.