What in the word?

Since headlines are designed to grab readers’ attention, kudos to the headline writer for a Sept. 29 New York Times op-ed that declared: “Writing Isn’t as Important as We Think.” Them’s fighting words for a museum of language arts! As thousands of readers probably did, I immediately started to read that provocatively headlined article.

Written by Williams College professor of philosophy Joe Cruz, it argued that writing is not an optimal way to develop or defend one’s ideas. Instead, he wrote, channeling Socrates, people should prioritize speech. Being able to muster arguments on one’s feet means, in Cruz’s view, that someone has a good grasp of material and is thinking straight.

Writing, he maintains, isn’t necessarily evidence of higher-order thinking. It is more akin to archiving (ouch!).

Not surprisingly, his op-ed provoked a large response (mostly negative, judging from the comments and letters to the editor that ran afterward). As one reader said, “Engaging fully in the writing process is, and should remain, a central tenet of liberal arts education. Why? Because it is hard. Perseverance is the point. Guiding a piece of writing from a bad first draft into a polished final product is a difficult but meaningful exercise for the mind.”

Writing is a technology invented by humans — early on, it served an important purpose in bookkeeping and trade. By recording their clay ledgers in cuneiform, our newly empowered ancestors could better track their accounts and the size of their herds. Thousands of years later, they would use their more advanced writing tools to record their culture’s stories and myths for posterity.

I guess one could plausibly argue that writing alleviated people from the task of memorization, but it doesn’t seem fair to then argue that writing also alleviates people from thinking, which is what Cruz seemed to be suggesting.

Why is a debate about the value of writing surfacing now? Probably because people are sincerely worried or confused about letting AI write for them. That seems to be an age-old pattern when new technologies arrive on the scene: Just as Socrates warned his pupils about writing and just as people in the Middle Ages were leery of the printing press, new technologies often scare people and deter them from becoming early adopters. They fear the effects of new technology and can’t imagine the new age of invention, creativity, and productivity that is to come.

But this time, I’d suggest, is different. Now we have the neuroscience and labs and tools that can measure whether we lose something by using this new technology, these AI writing assistants. And the results aren’t great. Scientists and mathematicians have already noticed that AI is relieving them from doing some of the grinding, difficult work that often leads to understanding and breakthroughs in their fields.

According to a June 2025 article in Education Week, electroencephalography (EEG) results “demonstrated that a group of writers who didn’t rely on outside support [chatbot help] had the strongest, widest-ranging neural activity.” The MIT Media Lab, which conducted the study, reported in its summary conclusion that:

Over four months, LLM users consistently underperformed at neural, linguistic, and behavioral levels. These results raise concerns about the long-term educational implications of LLM reliance and underscore the need for deeper inquiry into AI’s role in learning.

A large experiment carried out by a team at the University of Pennsylvania, cited in Education Week (May 28, 2026), found that students “who used [ChatGPT] demonstrated a 48% improvement in grades while working with an AI tutor, but when AI access was removed, they scored 17% worse than students who had never used AI at all.”  A Stanford University study confirmed this pattern.

What else would we stand to lose? Using AI for writing assistance can lead to a bland (some have said “robotic”) style and stilted vocabulary choice (there are already lists of words that AI is way more likely to use than humans). LLMs are still prone to hallucinations and bias and often rest on a foundation of misappropriated materials. Not to mention the serious environmental costs of all the data centers required to produce top-notch LLMs.

But since it’s a pretty good bet that people will continue to use AI to help them write, how should we think about it? Is it always bad, or are there times when it should be allowable assistive technology?

One interesting case came up in the national news this summer. The storied investor Stan Druckenmiller wrote an op-ed in the Wall Street Journal criticizing the decision of Secretary of the Treasury Scott Bessent to buy back Treasury bills in an attempt to lower the national debt. His piece was praised for being especially well-written, but it was subsequently revealed to have had help from a chatbot. Druckenmiller was then roundly criticized.

But he defended himself, saying that everyone knew where he stood on the issue of the national debt, having voiced the same warnings for years. Admitting to not being a great writer, he said he was glad to have had a chatbot assistant making his arguments more persuasive. That brought up a broader issue: Should he have acknowledged receiving help from AI, and if he had, would the WSJ still have run it or at least run it with a “double byline”?

In another controversial instance, cited by teacher Brett Vegelsinger in his book Artful AI in Writing Instruction, Google ran a commercial during the 2024 Olympics in which a father urged his daughter to ask Gemini for help composing a fan letter. That raised the question why. The charm of a fan letter is its authenticity — its expression of honest feelings that compelled someone to write.

Along the same lines, when we know a bot is behind the reams of texts we get from political candidates, aren’t we just a bit disappointed to realize that famous politicians haven’t actually reached out to us?

The New York Times interviewed a prolific author who goes by the pen name Coral Hart (and several others, depending on the genre she is producing) who submits outlines for her fantasy and romantasy novels to AI, which thus enables her to churn out some 200 books a year and garner a substantial, steady income. But she, like other authors who use AI, tends not to admit it. Says one romance book club leader, “Romance is about human connection, and there’s nothing more human than being vulnerable and falling in love. Why would I want to read a story written by a machine trying to emulate that?”

And what about using AI to write in classrooms? It seems to me that if writing and thinking and learning are inextricable, if we want to ensure that deep learning occurs, we must let students do the hard work of writing unaided, especially younger students. Perhaps AI shouldn’t be allowed in the classroom at all until students have mastered the fundamentals of reading and persuasive and creative writing on their own, just like calculators aren’t generally permitted until children have mastered the basic number sense that helps them avoid a crazy, probably mistyped math answer.

As we confront these types of pressing ethical and pedagogical questions in our new, unfamiliar, rapidly advancing technological era, we need a new set of rules to guide us.

Personally, I appreciate having an AI assistant do research for me — but only at my prompting and after I have independently come up with an idea. Then we all need to be sure to check any information a chatbot gives us for accuracy and sources.

I think AI should be acknowledged when we use it to substantially write a piece for publication, just like we would credit a co-author or even ghost writers. But most important, we must all keep reading and studying great literature and nonfiction narrative writing to learn to appreciate fresh imagery and style, to develop our own unique voices, to struggle with ideas and — to paraphrase Joan Didion — to find out what we think and, in so doing, to truly learn.

And that’s why writing is still as important as we think.

—Ann Friedman, founder of Planet Word