The 800-Word Book Review

The sentence seems to want to continue, "I begin to despair"

1. In the preface to Male Fantasies, a 1977 psychoanalytic study of “the fascistic type” in Nazi Germany, author Klaus Theweleit interrupts himself mid-sentence with an ellipsis and writes:

“...at this point, the sentence seems to want to continue, ‘I begin to despair’ (but this is the reality of a frozen semantics, not, in the end, of any feeling I might have).”¹

Theweleit then begins a new paragraph, leaving the previous sentence unfinished. When I read this passage, I was stunned. Think about what Theweleit is suggesting here: that he is not the writer of this text, but that the text is, in a sense, writing itself. The first half of the sentence naturally leads Theweleit’s pen to compose “I begin to despair,” but then he thinks, wait, I don’t actually think this—I’m not despairing—but I feel inclined to say it because of the structure of language (“a frozen semantics”).

Finally, I thought, a writer who is aware of the automaticity of language and who actively works against this, who realizes that we are not necessarily in control of language, but that language controls us, shaping the very thoughts and ideas we are able to have. For any writer, this is an unacceptable situation, which is why Theweleit must pause and then move in a new direction, taking back control of his words, but the fact that he leaves this “edit” in the text is remarkable. Rather than show us a smooth, polished piece of craft, he allows the imperfection to remain, a visible sign of his struggle to say something true (rather than regurgitated cliché) about fascism and Nazism.

Theweleit continues to do this throughout Male Fantasies, repeatedly critiquing theoretical and psychoanalytic language, which he notes can “reduce the phenomena under investigation to the most all-embracing of concepts possible.” In another formulation, Theweleit remarks that “conceptual constructions” often “do a better job of ignoring real phenomena than of abstracting them,” and that this constitutes a “widespread flight from concrete reality.” And yet, I wrote at the beginning of this essay that Male Fantasies is a “psychoanalytic study,” so why is Theweleit so insistent on critiquing psychoanalytic language and concepts?

Theweleit anticipates this question and responds by emphasizing the nature of his book, which “present[s] typical specimens of the writings of soldier males, sticking closely to the text in every case.” In other words, Male Fantasies, rather than starting from the viewpoint of psychoanalysis, or the position of the expert using theoretical tools to diagnose a pathology, begins from the beginning—from the memoirs of “soldier males” who made up the German Freikorps (roving paramilitary units who refused to disband in the aftermath of World War I) and who would go on to run the concentration camps and the gas chambers in Nazi Germany. Only by reading their words, Theweleit suggests, can we truly understand “the fascistic type,” whereas theories and concepts will simply obscure a truth we find too difficult to face. This truth, according to Barbara Enrenreich’s introduction, is that “the fascist is not doing ‘something else,’” as a result of, say, repressed desire or false consciousness, “but doing what he wants to do.” I thought of this when ICE—another paramilitary band operating outside of the structure of the law—was recently terrorizing and murdering civilians in Minnesota. Instead of looking for a hidden motive to their actions, instead of thinking that spreading terror and violence was really about “something else,” it was much more illuminating to accept the immediate, surface level explanation: terror and violence were the point all along.

2. I went searching for Theweleit’s line, “the sentence seems to want to continue,” and the remark about “frozen semantics,” which I’d read and remembered from months earlier, when I read about a writer who was discovered to have written part of a book review for The New York Times with the aid of AI. The writer was caught as his review borrowed liberally from a review of the same book in The Guardian, only minimally rephrasing certain sentences, and a reader had noticed the similarities. Apparently, the writer had not intentionally plagiarized the review in The Guardian, but the AI software had used it as a main source in the text it generated, making the connection obvious.

More than the simple use of AI, what intrigued me about this case was that the writer would use AI for a short, newspaper length book review of 800-1,000 words; in other words, that a writer would use AI to write an article that must conform to strict genre conventions, that must be, in some sense, formulaic. The 800 or so word book review, because of its limited length, follows certain guidelines: enough plot summary to give a reader an idea of the contents of the novel; some evaluation of its quality; a bit on the author’s biography and past work; and something on the novel’s form, style, or language. Once these boxes are checked, there’s little space left for the writer to include their own personal anecdotes, or to convey their phenomenological reading experience, or to wonder about how the form itself might be shaping their review, as in the example of Theweleit questioning his own language. Because of the rigidity of its form, the short book review—more so than the sprawling, idiosyncratic, and autobiographical review—is vulnerable to automation.

Last summer, a similar “scandal” occurred when the Chicago Sun-Times posted a “Summer reading list for 2025” that was discovered to have been generated by AI as it featured books that didn’t exist. We might ask the same question of this “listicle” as of the 800-word book review: why shouldn’t it be AI generated? The “listicle” is already a degraded form of writing, popularized as an easy way to drive traffic and generate ad revenue, and the 800-word book review can often seem indistinguishable from marketing copy, discussing the book in question with vague generalizations and clichéd language. To see this, one only has to read the original review in The Guardian that was plagiarized by AI, a 767-word article which refers to Italy—the setting of the novel in question—as “a resilient, self-sabotaging and remarkable nation” and then as “a country of contradictions, battered, war-torn, divided, misguided and miraculous.” There are even more adjectives used to describe Italy, but to me, these sentences tell us nothing, so I’m not going to include them.

I don’t want to be too hard on the writers of these reviews, however, for the same reasons that a criticism of the aesthetic qualities of a listicle would be misguided. Can one write a listicle that also functions as a compelling piece of criticism? Probably not. What about an 800-word review? Maybe, but as a writer, the deck is stacked against you. I criticized The Guardian review for including what I see as generic, boilerplate language, but with such limited space, the writer is encouraged by the form itself to use cliches and generalizations, as this is the surest way to cover the vast expanse of a novel. When the writer describes two characters and judges them, claiming that “contrary, self-sabotaging Mimo, ‘disagreeable’ by his own estimate, is a masterly creation, as is his proud, mercurial, impatient soulmate,” I am forced to take this assessment on faith alone, as there is not enough room for the writer to provide proof or evidence. But considering all that has been written about the elite, incestuous nature of publishing and reviewing, I’m simply not inclined to give someone my trust until they earn it through the review itself, and with only 800 words to work with, I’m not sure they can.

3. At this point, some readers might think I am advocating for the replacement of writers by AI in the book review sections of The New York Times and The Guardian, but this is not the case. Instead, I am advocating for writers to prevent themselves from becoming AI. This is what Theweleit is on guard against as well—whenever a writer feels that “the sentence seems to want to continue,” they must ask themselves if they are actually writing what they think, or if they are giving into “a frozen semantics.” A sentence continuing of its own accord is a perfect description of how ChatGPT functions—just think of the pulsing black dot as it “thinks” and the text that appears line by line, as if by magic; and “a frozen semantics” is also a compelling description for the generic AI voice that has become recognizable to many, relying heavily on certain words and sentence constructions (“It’s not X, it’s Y”). Considering this situation, writers attempting to say something meaningful and truthful must actively work against cliché, and tired forms must be renewed. But how? Even my call to “renew tired forms” is a cliché.

Two examples come to mind that provide a counterpoint to the 800-word book review. These are book reviews that, when I read them, stayed with me not necessarily because of the insight they provided into the book in question, but because they could stand alone as pieces of writing worth reading on their own. The first is Karl Ove Knausgaard’s review of Michel Houellebecq’s Submission in The New York Times from 2015. Knausgaard begins his review by admitting that he’s never read Houellebecq. Then he gets to the first sentence of Submission, which mentions the 19^th century French writer Joris-Karl Huysmans, and he mentions he hasn’t read Huysmans, either, so he decides he’s going to read Huysmans to better understand Submission:

I could probably have read “Submission” with that image in mind, and then looked up Huysmans on Wikipedia before embarking on this review, but something about Houellebecq’s use of Huysmans struck me as fundamental to the novel, it seemed almost as if the ambition, in a way, was to rewrite Huysmans, to test out his conflicts in our day and age, to create a sounding board of the kind only novels can create, and so I took Huysmans’s best-known book, “Against the Grain,” with me to my daughter’s gymnastics practice and sat on the benches, drinking coffee from a plastic cup and reading while she somersaulted about on the mats below along with perhaps a score of other 10-year-old girls, in a harsh and glaring light as one hit song after another blared out of the public address system.

What a sentence. Here Knausgaard is not giving us his assessment of Submission, as one might expect in a book review; instead, he’s recreating his experience of reading Submission, which starts by reading Huysmans. As he says, he could have “looked up Huysmans on Wikipedia,” the equivalent of which in 2026 would be asking ChatGPT about him, and he could have concealed this fact, simply dropping some information on Husymans into the review, making himself appear as an expert; however, Knausgaard begins from a position of naïveté and, like Theweleit, he begins at the beginning. This leads to his review being insightful and allows it to get to the core of the novel; Knausgaard realizes that Submission’s scorn is directed at “the intellectual classes” of France, while “the Islamization of France,” which many in the media took to be the book’s main theme, “is merely a consequence” of the West’s spiritual emptiness—the same emptiness that has been Houellebecq’s primary concern throughout all his novels. Houellebecq’s target is himself, his own culture, and his own world.

The other review I have in mind is a more recent one from Michael Hofmann on the novels of Andrey Platonov. Hofmann is somewhat famous for his highly personal and critical reviews, and in this one, published in the New York Review of Books, he spends the majority of the review explaining how difficult it was for him to read Platonov and thus, write the review. I could pick any number of passages to quote, but consider this one:

Reading Platonov is like following a fire hose hundreds of miles long, held by thousands of anonymous—though named—firemen. It is like encountering a novel written in haiku. It has smooth ball bearings for clockwork. There is no grip, no engagement, no involvement. And accordingly, I might have been my small granddaughter, complaisantly holding the books upside down. Weeks, months later, I still felt a deep unwillingness to open them.

Here Hofmann is doing something similar to Knausgaard; rather than judge from on high, he brings us into the reading experience itself, and in doing so, says something interesting about Platonov and his style. How curious, too, that both writers mention young children, as if they are implying that to really understand a novel, you have to let yourself become a child again, you have to experience it in an immediate way, and only afterwards can you step back, assess, and evaluate. We don’t have to agree with Hofmann on Platonov, just like we don’t have to agree with Knausgaard on Houellebecq, but they’re not asking us to agree with them. They’re saying, “This is what reading this book was like for me.” And in this, the reader feels welcomed, invited into the reading experience, so that they can answer the question: “Is it like that for you, too?” The book review becomes literature in the way that Kazuo Ishiguro described it:

[S]tories are about one person saying to another: This is the way it feels to me. Can you understand what I’m saying? Does it also feel this way to you?

AI can’t feel. Humans can, but we can also ignore our feelings, and we can refuse to think, to consider, and to imagine deeply. We can write careless sentences that are passable but uninspired, and we can churn out short reviews that sound like AI. Knausgaard didn’t have to read Husymans, and Hofmann didn’t have to “lug them [Platonov’s novels] back and forth across the Atlantic,” but they did, and it’s evident in their reviews. For now, the personal and the phenomenological remain the domain of humans, but they have to be cultivated and nurtured because the world, as always, threatens to turn people into things. Book reviews, if they should continue to exist, should not be lifeless things, but should exhibit all the capacities of their writer. They might feature the reading experience, as Knausgaard and Hoffman do, but on a broader level, they should exhibit an awareness of form, and the writer should either elevate the form or break it open, rather than simply reproduce it.

4. I cheated a little bit in mentioning the Knausgaard review (4,500 words) and the Hofmann review (2,500 words). In the 800-1,000-word review, the writer doesn’t have the luxury of spending 1,000 words on a different book, like Knausgaard does, or of complaining for 1,000 words, like Hofmann does. So what is the short newspaper length review even for?

In a 2022 article on the history of The New York Times Book Review, Kyle Paoletta considers this question by tracing the way the Book Review’s focus has shifted throughout its long history (over 125 years), going back and forth between a focus on “books as news” and a more critical sensibility. He highlights the creation of The New York Review of Books (in direct opposition to the Book Review) in 1963 as a seminal moment, spearheaded by Elizabeth Hardwick and her 1959 article “The Decline of Book Reviewing,” which criticized the Book Review for, among other things, existing “in a state of baneful depression insofar as liveliness and interest are concerned.”² My complaints, then, are nothing new. The short book review has been criticized since the beginning for functioning more as publicity than as criticism; in Christian Lorentzen’s 2019 article, “Like This or Die,” which covers similar historical territory as the Paoletta article, Lorentzen also comments on how recent short reviews have functioned more as recommendation than as criticism. Both articles show that the short book review was never intended to be pure criticism but was always tied up with marketing, publicity, and business. Acknowledging this, we are returned to my original question: why shouldn’t they be generated by AI?

If an advertisement was discovered to have been created by AI, there wouldn’t be an uproar, because advertising is simply a means to an end, whereas art is an end in itself. Criticism, most writers and critics would probably agree, is an art; but what about the short book review? It’s part criticism, part publicity, and perhaps because it sits in this strange middle ground, the AI review in The New York Times has caused such controversy.

At the end of her essay, Hardwick calls for “the unusual, the difficult, the lengthy, the intransigent, and above all, the interesting” in book reviewing. I agree, but I also don’t want to simply do away with the short book review. The long reviews that I prefer, which are featured in literary magazines that come out weekly or monthly, have the benefit of depth, but as Paoletta points out, “The Book Review’s greatest strength…has always been the breadth of material that it’s able to sift through and consider in a year.” The New York Times and The Guardian, to focus on the two publications where the reviews related to the AI controversy were published, put out new reviews every day. As far as I’m aware, no other major publication does this, and The Washington Post recently closed its book section, “Book World.”

But rather than just lamenting the cuts to books coverage, as many did in the case of The Washington Post (myself included), the newspaper length book review should be written as if it actually matters. This would mean, even in a short space, guarding against cliché, avoiding generalizations, including oneself in the writing process, and considering the form of the short review, among other things. I’m not saying it’s easy, but the effort should be made. As a follow up to this essay, for the next book I read, I’ll write a 800-1,000 word review to see what I can do. In the back of my mind, I’ll have Theweleit’s words, reminding me to pause whenever the sentence wants to continue on its own, and to consider if what I’m writing has worth, or is just the result of a frozen semantics, a form that has outlived its usefulness. ■

1. The full sentence reads: “When she tells me of contact, true contact with lightning and rolling thunder, of the production of an intimacy through the feeling and exploration of distances—an intimacy that is not consuming, distance that is not far away—a place where caution is a beautiful word, related to foresight and to a feeling for the reality of a suffering that wishes for change but is caught in impasse and double bind, I think then of the studied or hectic nonchalance of all those (myself included) who are striving to combat fascism here and now, but are blind to the experience of the nonfascist . . . at this point, the sentence seems to want to continue, “I begin to despair” (but this is the reality of a frozen semantics, not, in the end, of any feeling I might have).”

2. Hardwick quotes lines from five different book reviews in her article that are extremely generic and vague. If they had been written today, they might be accused of having been written by AI.