How will AI shape criticism?

Criticism no longer begins with human readers.

When a book is published, it now enters a world in which machines are among its earliest interpreters. It is subject to AI-generated summaries not by people who have read it, but by systems trained on vast quantities of the internet and other open sources.

Shortly after my own recent book — Dr Bot: Why Doctors Can Fail Us and How AI Could Save Lives — appeared, Google’s very own bot produced a neat summary of it. The AI also confidently attributed authorship to a man called Stuart Blakley. I do know Stuart Blakley — he is a good friend who wrote a single blog post about my book. Despite a CNN book appearance, a review in the Washington Post and an academic career spanning nearly twenty years, AI could not come to terms with the fact that a female author wrote a popular science book.

Around the same time, a generous “review” arrived from an academic at an Ivy League institution which read uncannily like another reviewer’s prose, gently rearranged and repackaged. The sentiments were also generous, but the rhythm of the writing was assembly-line: the sentences were short, choppy and vague. And while it was gratifying to receive praise, I suspected this person had not cracked the book’s spine. The review suggested a volume whose internal architecture I barely recognised.

None of this was terribly surprising to me. When you get lemons, as the saying goes, make lemonade. I duly made some fizzy social media posts, pointing out the chapters of my book devoted to AI bias and “hallucinations” —  bots’ tendencies to make up stuff — outputs which look slick but can be woefully incorrect.

The experience also felt like an early glimpse of a broader shift: AI not merely as an object of criticism, but as a participant in it — shaping how books are summarised, culturally framed and in some cases, like my own, effectively reviewed before human critics have done their own due diligence.

Cultural criticism is not the only place in which AI is creeping. It is visible in plenty of other intellectual and knowledge-based fields. Take healthcare: significant numbers of clinicians and patients alike use generative AI tools — predominantly, my research has found, the consumer tool “ChatGPT”. In the UK, in anonymous surveys, at least a quarter of general practitioners say they are using these kinds of AI. 

Face to face, however, neither party — doctors nor patients — tend to be terribly candid about this. We’ve experienced this trend before. One Harvard physician told me that he left patients in the office when he wanted to Google their symptoms. Similarly, with these easier to use, “chattier” AI tools, doctors fear they’ll be considered “cheats” if they disclose what they’re up to. Out of social etiquette and fear, patients prefer not to use AI to question or challenge higher status medical authorities – after all, our lives are still in human doctors’ hands.

The role of the arts and books critic is considered special. Owing to his or her self-image as an arbiter of culture, the critic — like the artist — may be more embarrassed than others to confess any possible use of AI tools. Such a reflex is traceable to the old “two cultures” war — the idea that science is the domain of instruments while art is the ineffable domain of humanity. Yet culture has always been produced through tools and technologies — from the crude to the sophisticated. What has changed with AI is the scale speed, and the ease with which artistic mediation and creativity can arise.

Anecdotally, one can quite easily discern some clear-cut cases of AI adoption. Several of my colleagues’ emails became, almost overnight, more fluent, more polite and more Californian in cadence. Similarly, on LinkedIn and other professional platforms, a recognisable style emerged: staccato sentences, breathless yet bland enthusiasm with bullet points and em dashes aplenty.

Conceivably, more sophisticated AI use could lead to a gradual reconfiguration of artistic skill

However, not all adoption is easy to spot. Generative AI is not only capable of producing text; it can help craft what “works”. These tools can engage in “dialogue” with the writer: new ideas can be thrown back in seconds, a paragraph can be proposed in alternative styles and metaphors tested and discarded. Used in this way, AI is less banal ghostwriter than a stylistic companion. This may well allow many writers to produce cleaner, more elegant prose.

Conceivably, more sophisticated AI use could lead to a gradual reconfiguration of artistic skill. For example, the question of who is wholly responsible for a thought, a crafted sentence or an idea begins to blur with these tools. Feedback, drafting and discarding have always been essential in creating good art. This process could be expedited — at least in part — with AI. Naturally, there is a worry that something is lost. But creativity may not be threatened — instead, it could arise via new means and methods. The romantic idea of the tortured artist may wane but then again, is suffering for one’s art always inevitable? Perhaps the artist could toughen up and just use technology.

Some are honest about this. In 2024, the Japanese novelist Rie Kudan confessed that around five percent of her prize-winning novel had been generated with ChatGPT. I recently asked a London publisher friend to gauge what percentage of his writers used generative AI tools. Without so much as a blink, he said, “all of them.”

Again, one can observe this clearly in other fields which are acknowledging AI’s role. Consider medicine again. The New England Journal of Medicine AI is now experimenting with “AI” as a peer reviewer alongside human experts. While some medical academics may appear dramatically aghast at this, privately, they may find the development unsurprising. Most academic journals, after all, now ask the question about how AI is used in manuscript preparation. In one paper that I recently reviewed, the authors reported: “Generative AI was used for synonyms”. I strongly suspected, however, the tools played a rather grander role than as a digital thesaurus.

Current debates on AI focus on writers and critics as the injured party in the rise of generative AI, a community whose labour has been appropriated and whose artistic integrity and future has been placed in jeopardy by systems trained on their original work. There is truth in this, and, as a writer, I have skin in the game. But the truth is selective, and there is more to the story. Many of the same people articulating these anxieties are, quietly and without any overt ceremony, incorporating these tools into their own creations. Equally interesting is the question of how a reader distinguishes between writing solely created by humans, that co-written with technology and prose completely outsourced to it. 

The Critic excludes AI-generated articles, but it is worth considering the ongoing relationship between AI and human creativity. Did I use generative AI to write my own book? No — which may help to explain why it took the better part of a decade to pull it together. Would I use it to assist me in the future? Reader, almost certainly.