Posted on The Velvet Quill by KL Adams

On January 6, 2026, the New York Times published a book review of Watching Over Her by Jean-Baptiste Andrea. The reviewer was Alex Preston — a British author and journalist who had written for the Financial Times, the Economist, the Guardian, and Harper’s Bazaar. A credentialed, established, respected literary critic.

The review ran. Readers read it. Life went on.

Then, in late March, a reader noticed something.

Preston’s review contained language nearly identical to a review of the same novel published in The Guardian in August 2025 — four months earlier — by critic Christobel Kent. Not similar in argument or approach. Similar in words. Specific adjectives. Structural observations. Character descriptions lifted almost verbatim.

Kent had written: “lazy Machiavellian Stefano.”

Preston had written: “lazy, Machiavellian Stefano.”

The reader alerted the Times. The Times investigated. Preston admitted he had used an AI tool to help draft the review. The tool had gone online, found Kent’s Guardian piece, and incorporated passages from it into Preston’s draft. Preston hadn’t caught it. He submitted the piece. The Times published it.

The Times severed ties with Preston immediately. An editor’s note now appears at the top of the published review. Preston told the Guardian he was “hugely embarrassed” and had “made a serious mistake.” He apologized to the Times, to the Guardian, and to Christobel Kent personally.

None of that undoes what happened.

What Actually Went Wrong

The Preston case is being covered primarily as an AI story — a cautionary tale about what happens when a writer uses a tool they don’t fully understand. And it is that. But it is also three other things that are more interesting and more uncomfortable.

It is a story about invisible plagiarism. Christobel Kent wrote a book review. It was scraped into an AI tool’s training or retrieval system without her knowledge or consent. That tool then reproduced her words in a publication that fired her colleague — the tool itself — for copying her. Kent was neither credited nor compensated. Her name does not appear in the editor’s note. The Times’ apology was not to her. She was simply the source material.

It is a story about editorial oversight. Preston had written six reviews for the Times since 2021. Someone at the Times read this piece before publication. Editors exist. Fact-checkers exist. The similarities to Kent’s Guardian review were specific and structural — identical adjectives, matching metaphors, nearly verbatim conclusions. Nobody caught it. That is not Preston’s failure alone. That is a systems failure.

It is a story about what we actually mean when we say “I wrote this.” Preston told TheWrap he used “an AI editing tool improperly on a draft I had written” and that he failed to catch “overlapping language” from the Guardian review. His explanation implies he wrote the review first and used the AI tool as an editor. The tool apparently did more than edit. The question of where Preston’s draft ended and the tool’s retrieval began is genuinely unclear — and that murkiness is the whole problem.

The Reviewer’s Responsibility

Literary criticism is not simply opinion. It is a specific act with specific obligations.

When a reviewer reads a book, sits with it, and writes about it, they serve as an intermediary between the author who created it and the reader who is deciding whether to engage with it. The review is a record of one person’s genuine encounter with a piece of work. That encounter cannot be outsourced.

An Australian literature academic described it precisely: “When we write reviews, we have to do it naked — as individual readers, with a public to judge our judgments.” The reviewer sits at the center of a pact between writer and reader. Done well, criticism is itself a form of literature.

What AI does to that pact is not simply to introduce efficiency. It substitutes someone else’s encounter for your own. Preston may have read Watching Over Her. He may have formed real opinions about it. But the review that was published incorporated another critic’s words and observations — Kent’s specific responses to specific passages — as if they were his own. The reader who read that review was not getting Preston’s encounter with the book. They were getting a composite that neither Preston nor the Times fully controlled.

That is not a minor lapse. It is the entire point of the job, and it failed.

What Happened to Christobel Kent

This is the part of the story that is not getting enough attention.

Christobel Kent wrote a book review. It was good enough for an AI tool to use it as a source. Her words were specific enough that they transferred almost verbatim into another piece. Her work was used without attribution, without compensation, without so much as a mention in the subsequent scandal.

The editor’s note in the Times appended to Preston’s review links to the Guardian piece. Kent’s name appears there. That is the full extent of the acknowledgment she received.

She did not choose to become a story about AI in publishing. She did not consent to her work being used as training or retrieval data. She had no recourse. She received no apology from the tool that copied her. She received no payment for the use of her work. She eventually received a hyperlink.

This is the part of the AI plagiarism conversation that needs to be louder. The Shy Girl scandal focused on a publisher canceling a book. The Preston scandal focused on a journalist being fired. In both cases, the collateral damage — the author whose prose was questioned unfairly, the critic whose words were stolen quietly — received a fraction of the attention.

The NYT’s Response

In the aftermath of the Preston case, the Times issued a “periodic reminder” to all its freelancers about AI policy. The memo was unambiguous: all writing submitted to the Times must be “the product of human creativity and craft.” Freelancers “may not use generative AI tools to help you write any part of a story.” Using AI tools to “create, draft, guide, clean up, edit, improve, or rephrase your writing is strictly prohibited.”

The specific tools named in the prohibition include Gemini, Claude, ChatGPT, and Perplexity. Image generators were also prohibited.

The memo came, notably, after a string of AI controversies at the Times — not before them. It is a reactive document, not a preventive one. The stable door closed after the horse had already been published in the Book Review.

The harder question is what the Times’ internal use of AI looks like. The memo covers freelancers specifically. In-house journalists, per the Times’ own statement, have “separate guidelines for using AI and approved GenAI tools.” What those guidelines permit is not public.

What This Week Taught Publishing

The Preston case and the Shy Girl cancellation landed within weeks of each other. Together, they exposed something the industry had been managing to avoid looking at directly: there is no consistent standard, no reliable enforcement mechanism, and no coherent ethical framework for AI use in publishing.

Publishers are canceling books for suspected AI use, relying on detection tools that universities have banned as unreliable. Critics are being fired for using AI tools they didn’t fully understand. Authors are being publicly accused by a detection company whose CEO names individuals on social media. And somewhere in all of it, the people whose actual work was used without permission — scraped, incorporated, reproduced — are being treated as footnotes.

The publishing industry spent years building trust with readers on the premise that the words in a book, a review, a column, represent a human mind genuinely engaging with ideas and stories. That premise is not dead. But it is under serious pressure.

The reader who noticed the similarities between Preston’s review and Kent’s — who read carefully enough to catch what the Times’ editors missed — is the best argument for why this all matters. Readers notice. Readers care. Readers are, in the end, the ones holding the industry accountable when everyone else has stopped paying attention.

That reader is why we have to keep talking about this.

What do you think should happen when a tool creates plagiarism that the writer didn’t intend? Where does responsibility sit — with the writer, the tool, or the publisher? Tell me in the comments.

KL Adams is a literary blogger and fiction writer specializing in dark fantasy, vampire fiction, and paranormal romance. Follow on WordPress, Inkitt, and Wattpad for reviews, reading lists, and stories that haunt you long after the last page.

📚 Building your own dark fantasy world? Grab the Dark Fantasy Worldbuilding Kit on Gumroad — worksheets, magic system builders, character prompts, and storytelling tools designed specifically for dark fantasy, gothic fantasy, and romantasy writers.