The internet is a towering babel of disposable language.

It is strange to be a writer on the internet these days.

In the 1980s when I was an English major at a state college in New Jersey, I imagined a future in which authorship and craft would be highly valued—an information ecosystem dominated by books, journals, and magazines—but the internet would ultimately create a much different tableau for the written word. A large percentage of internet writing is unauthored, and much of it is presented in formats that prevent readers from clearly understanding where it originated. From its infancy, the internet encouraged anonymity in BBSs and Usenet groups where nicknames or pseudonyms were commonplace. Concerns about online privacy and the ease of concealing one’s identity online contributed to this erosion of traditional authorship. The internet is also full of duplicated and plagiarized content. A research study published in 2015 by the digital marketing company Raven Tools found that of 200 million pages analyzed, 29% had duplicate content.

In my journalism and creative writing classes, I was taught that a writer tries to meaningfully connect with readers. You can do this on the internet, but so much online writing is created for less elevated purposes. "Clickbait" articles, for instance, present content under sensationalist, emotionally charged headlines to pump up site traffic, engagement, and ad revenue, often yielding less than 30 seconds of actual reading time. The same can be said of terabytes of forgettable third-party content that streams by on social media feeds every day. SEO, an entire industry dedicated to "optimizing" content within search engines, tirelessly generates verbiage explicitly to be read by algorithms. The internet is also full of language that is the by-product of online activities, the "digital exhaust" of social media activity, browser history, transaction records, and data from smart devices.

As a young man, I just assumed that humans would always be the writers, in the linguistic driver's seat. I was wrong. Online writing is increasingly automated. E-commerce companies deploy AI to push out product descriptions, and SEO and marketing companies use AI to turn the crank on content creation.  Even reputable news organizations such as the Associated Press, The Washington Post, Bloomberg, and Reuters automate content now. AI-generated content is turning up everywhere online like an algae bloom choking out other life forms in a lake. A 2024 investigation by Wired Magazine determined that 40-47% of new posts on Medium were written by AI. Literary magazine editors now regularly complain about being deluged by AI-written short stories. In 2023, NewsGuard uncovered 49 websites churning out entirely AI-generated clickbait in Chinese, Czech, English, French, Portuguese, Tagalog, and Thai.

In 2020, an SEO toolset company called Arefs researched internet traffic on 14 billion web pages. Among other things, they discovered that 96% had no organic search traffic at all. This means that most online content, including most of its writing, is invisible to internet users.

This is the perfect illustration of our emergent digital civilization. On the surface, the internet presents with dynamism and liveliness, insisting that it will encompass all life someday and making frantic demands for our attention, like a needy child, but just underneath lies a vast becalmed ocean of dead language.

Like a landscape littered with old factory smokestacks, the internet of dead language has been belching out low-stakes writing for decades. When I first heard the term “content farm” fifteen years ago, Demand Media was publishing 4,000 articles per day, more than 1 million each year. EHow, Demand’s top-performing content farm, had more than 2 million articles. Companies like Associated Content, Braxton, and ZergNet also flooded the zone with short, forgettable articles whose only purpose was to generate a few seconds of engagement online. This was the modern-day advertorial fired out of a machine gun that never clogs and keeps shooting indefinitely.

We writers knew intuitively that the hackers and disruptors would come for language eventually, and they finally have, gleefully and full of hubris, declaring themselves saviors in the process. There is no denying that this latest generation of generative AI is a remarkable technical achievement, but it was only possible in an ecosystem that had already devalued writing.

Writers have had a front-row seat to this devaluation. The new generative AI chatbots were created by looting the creative IP of writers and artists to train large language models. OpenAI used tens of thousands of published books and possibly hundreds of thousands shorter works in this training, pirating this content for their purposes. This activity prompted several lawsuits. In one ongoing class-action lawsuit, a group of artists including Sarah Andersen, Kelly McKernan, and Karla Ortiz have accused Stability AI, Midjourney, and DeviantArt profited from their work without their consent.

Writers have also witnessed this devaluation in the marketplace for writing and editing services. Fifteen years ago, freelance writers were being paid $3 - $15 per article for pieces that took 1-2 hours to write, rates that prompted justifiable accusations of exploitation. The compensation model has not improved, and AI is pushing the pay downward. Many of the job ads on freelance platforms are seeking people to edit AI-generated content rather than write it, offering $2–$10 per piece. Everywhere writers turn, they find companies and organizations using AI to do what writers and editors used to do. In 2023, for example, BuzzFeed laid off 180 workers, 15% of its staff, and shut down BuzzFeed News, while also announcing that it will use AI to create more content. AI has proven quite adept at writing forgettable summaries of celebrity gossip news, soulless (literally) product reviews, inane listicles, and brand narratives.

And the indignities continue to mount. The job market for freelance writers is now overrun by ads for AI training jobs that promise writers a paltry $20 per hour to train their AI replacement, with no benefits and the promise of eventual intractable unemployment. These ads have a respectable veneer, as if “AI trainer” is a job that has existed for decades, but underneath I detect a message that can only be seen through my writerly X-ray glasses:

Words are data. We own you.

With hindsight, it was inevitable that the internet would become a dead letter office for civilization’s long romance with writing. The old writing technologies imposed limits on language production. Only so much paper could be made from mulberry bark in ancient China; we should not be surprised then to learn that scribes were valued members of society and calligraphy an art form. The printing press democratized writing, but it also established the page with its precise geography and clear boundaries as the place where writing lived. One could only write so fast with a typewriter or pencil, and these limits imbued the act of writing with reflective, meditative qualities. Even a laptop has baked-in limitations because the pathway from an idea to fingertips to a screen is imperfect and fraught with the risk of disruption or outright failure. But Moore’s Law kept blowing out the capacity for memory and processing power, and the internet kept inviting language into itself. We kept filling it with words, frenetically, one dumb clickbait article at a time.

And more recently, generative AI has removed craft from the equation. There no longer needs to be a writerly "process" behind writing. All that tiresome, tedious work of pondering words, writing and rewriting sentences can be avoided. Now, you can simply turn on the syntax generator and point the hose in the general direction of an audience.