The 20th-century novelist and critic Cyril Connolly only ever wrote one good book, The Enemies of Promise, which was an attempt at understanding why he never wrote a great one. Granted, there was the booze and the sex and the Eton education. But, he said, the biggest factor was his choice of profession: ever since the age of 23, he had reviewed other people’s books.

He illustrates this with a fictional example: Walter Savage Shelleyblake is a writer asked to review books following the minor success of his debut novel Vernal Aires. Soon, Shelleyblake finds himself in a predicament: he must review books so that he can afford to write more books, but if he only ever reviews books, then he won’t have time to write them.

Today’s novelists no longer need to waste all their energy on deadlines. That’s because they have AI, which enables them to focus on the real stuff. At least, that might explain the English writer Alex Preston’s ruinous decision to use ChatGPT for a book review for the New York Times. Preston was asked to write 1,000 words on Jean-Baptiste Andrea’s Watching Over Her, with the review published on 6 January. It took almost three months before anyone alerted the NYT to the fact that Preston’s review had been written with help from a chatbot. But it was also flagged that he had repeated almost verbatim several phrases and passages from Christobel Kent’s Guardian review of that same book last summer.

But why do it? Preston told Sam Leith, the Spectator’s literary editor, that he had “written a draft review of the book, but it was under length, and I was rushing badly and drowning slightly. I made the stupid decision to use an AI tool to help expand and smooth it, with instructions about US spelling and house style at the NYT.” Putting aside how unusual it is for a journalist to find themselves under a word limit, or to worry quite so much about Americanizing their spelling, there will always be anxious writers like Preston, stumbling over deadlines with a chatbot crutch. The only surprise is that it was a book reviewer who did this, and book reviewers are supposed to be writers — the sort of people who care about writing, and so shouldn’t delegate that arduous task to an LLM.

Last summer, the Times columnist James Marriott was mocked for a tweet that shared an AI-written review. He suggested “you could submit… to a magazine and get published”, but he was only half-vindicated. That is a pastiche of a more adjective-swaddled era of reviewing, whereas the Preston scandal shows how reviewing has become less about style than about summary, less about craft than content.

Often mocked in their former pomp, the books pages used to be the only point of contact between literary culture and the ordinary reader. But reviews no longer seek to draw in outsiders: nowadays, they’re often indistinguishable from press releases. The only person I have known who bought books based on book reviews was my mother. She stopped last year for that reason.

Reflecting on his discussion with Preston, Sam Leith suggested that AI-assisted reviews are unlikely to happen too often. “There’s the practical point that by the time most critics file book reviews… there won’t be many or any other reviews to plagiarize from.” But the real issue isn’t AI panic — it’s how bland and uninspiring book reviewing has become.

---

Cosmo Adair is an editorial assistant at UnHerd.

cosiadair

---