A perfect storm is brewing for reading.
AI arrived as both kids and adults were already spending less time reading books than they did in the not-so-distant past.
As a linguist, I study how technology influences the ways people read, write and think.
This includes the impact of artificial intelligence, which is dramatically changing how people engage with books or other kinds of writing, whether it’s assigned, used for research or read for pleasure. I worry that AI is accelerating an ongoing shift in the value people place on reading as a human endeavor.
Everything but the book
AI’s writing skills have gotten plenty of attention. But researchers and teachers are only now starting to talk about AI’s ability to “read” massive datasets before churning out summaries, analyses or comparisons of books, essays and articles.
Need to read a novel for class? These days, you might get by with skimming through an AI-generated summary of the plot and key themes. This kind of possibility, which undermines people’s motivation to read on their own, prompted me to write a book about the pros and cons of letting AI do the reading for you.
Palming off the work of summarizing or analyzing texts is hardly new. CliffsNotes dates back to the late 1950s. Centuries earlier, the Royal Society of London began producing summaries of the scientific papers that appeared in its voluminous “Philosophical Transactions.” By the mid-20th century, abstracts had become ubiquitous in scholarly articles. Potential readers could now peruse the abstract before deciding whether to tackle the piece in its entirety.
The internet opened up an array of additional reading shortcuts. For instance, Blinkist is an app-based, subscription service that condenses mostly nonfiction books into roughly 15-minute summaries – called “Blinks” – that are available in both audio and text.
But generative AI elevates such workarounds to new heights. AI-driven apps like BooksAI provide the kinds of summaries and analyses that used to be crafted by humans. Meanwhile, BookAI.chat invites you to “chat” with books. In neither case do you need to read the books yourself.
If you’re a student asked to compare Mark Twain’s “The Adventures of Huckleberry Finn” with J. D. Salinger’s “The Catcher in the Rye” as coming-of-age novels, CliffsNotes only gets you so far. Sure, you can read summaries of each book, but you still must do the comparison yourself. With general large language models or specialized tools such as Google NotebookLM, AI handles both the “reading” and the comparing, even generating smart questions to pose in class.
The downside is that you lose out on a critical benefit of reading a coming-of-age novel: the personal growth that comes from vicariously experiencing the protagonist’s struggles.
In the world of academic research, AI offerings like SciSpace, Elicit and Consensus combine the power of search engines and large language models. They locate relevant articles and then summarize and synthesize them, slashing the hours needed to conduct literature reviews. On its website, Elsevier’s ScienceDirect AI gloats: “Goodbye wasted reading time. Hello relevance.”
Maybe. Excluded from the process is judging for yourself what counts as relevant and making your own connections between ideas.
Reader unfriendly?
Even before generative AI went mainstream, fewer people were reading books, whether for pleasure or for class.
In the U.S., the National Assessment of Educational Progress reported that the number of fourth graders who read for fun almost every day slipped from 53% in 1984 to 39% in 2022. For eighth graders? From 35% in 1984 to 14% in 2023. The U.K.’s 2024 National Literacy Trust survey revealed that only one in three 8- to 18-year-olds said they enjoyed reading in their spare time, a drop of almost 9 percentage points from just the previous year.
Similar trends exist among older students. In a 2018 survey of 600,000 15-year-olds across 79 countries, 49% reported reading only when they had to. That’s up from 36% about a decade earlier.
The picture for college students is no brighter. A spate of recent articles has chronicled how little reading is happening in American higher education. My work with literacy researcher Anne Mangen found that faculty are reducing the amount of reading they assign, often in response to students refusing to do it.
Emblematic of the problem is a troubling observation from cultural commentator David Brooks:
“I once asked a group of students on their final day at their prestigious university what book had changed their life over the previous four years. A long, awkward silence followed. Finally a student said: ‘You have to understand, we don’t read like that. We only sample enough of each book to get through the class.’”
Now adults … according to YouGov, just 54% of Americans read at least one book in 2023. The situation in South Korea is even bleaker, where only 43% of adults said they had read at least one book in 2023, down from almost 87% in 1994. In the U.K., The Reading Agency observed declines in adult reading and hinted at one reason why. In 2024, 35% of adults identified as lapsed readers – they once read regularly, but no longer do. Of those lapsed readers, 26% indicated they had stopped reading because of time spent on social media.
The phrase “lapsed reader” might now apply to anyone who deprioritizes reading, whether it’s due to lack of interest, devoting more time to social media or letting AI do the reading for you.
All that’s lost, missed and forgotten
Why read in the first place?
The justifications are endless, as are the streams of books and websites making the case. There’s reading for pleasure, stress reduction, learning and personal development.
You can find correlations between reading and brain growth in children, happiness, longevity and slowing cognitive decline.
This last issue is particularly relevant as people increasingly let AI do cognitive work on their behalf, a process known as cognitive offloading. Research has emerged showing the extent to which people are engaging in cognitive offloading when they use AI. The evidence reveals that the more users rely on AI to perform work for them, the less they see themselves as drawing upon their own thinking capacities. A study employing EEG measurements found different brain connectivity patterns when participants enlisted AI to help them write an essay than when writing it on their own.
It’s too soon to know what effects AI might have on our long-term ability to think for ourselves. What’s more, the research so far has largely focused on writing tasks or general use of AI tools, not on reading. But if we lose practice in reading and analyzing and formulating our own interpretations, those skills are at risk of weakening.
Cognitive skills aren’t the only thing at stake when we rely too heavily on AI to do our reading work for us. We also miss out on so much of what makes reading enjoyable – encountering a moving piece of dialogue, relishing a turn of phrase, connecting with a character.
AI’s lure of efficiency is tantalizing. But it risks undermining the benefits of literacy.
Naomi S. Baron does not work for, consult, own shares in or receive funding from any company or organization that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.
This article was originally published on The Conversation. Read the original article.
