AI vs writers: Is human writing under threat?
Writing has never been easy. Franz Kafka described it as a process of “unending torments”, while George Orwell compared it to “a long bout of some painful illness”.
Now, artificial intelligence promises to remove much of that struggle.
Large language models can produce emails within seconds, generate competent articles and essays and even attempt poetry. Writer’s block, once a familiar occupational hazard, means little to a chatbot capable of producing thousands of words on demand.
For human writers, however, the AI revolution is becoming increasingly alarming.
The scale of adoption is already significant. The UK Government’s latest public engagement survey found that 59% of adults had used generative AI in the previous three months, with 56% using AI capable of generating human-like text or speech. Among generative AI users, 58% said they had used it to write or edit documents.
For professional writers, the anxiety is even more pronounced. A 2026 ProCopywriters survey found that 74% of UK copywriters now use generative AI at work, up from 59% two years earlier. Yet 64% said they would press a hypothetical button to remove generative AI from the world, while 35% said they were obliged to use AI by a client, colleague or employer.
The Society of Authors has also found that 65% of fiction writers and 57% of non-fiction writers believe generative AI will negatively affect their future creative income. More than a third of translators surveyed said they had already lost work because of AI.
The concern is not simply that machines can produce words. It is whether businesses will continue paying humans to spend hours researching, thinking, reporting and crafting those words when AI can generate acceptable copy almost instantly.
There is also an emerging crisis of trust. Writers who use AI may face accusations of taking shortcuts, while those who do not may find their work wrongly flagged as machine-generated. Recent debate around AI-detection tools has highlighted the danger of false accusations, demonstrating how difficult it is becoming to establish who, or what, actually wrote something.
Yet writing is about more than assembling grammatically correct sentences. Human writers bring lived experience, judgement, emotion, cultural understanding, curiosity and, crucially, accountability.
AI may make writing faster. But if speed becomes more valuable than thought, the industry risks losing something far more important than jobs: the human voice.
Asian Voice spoke to experts to examine where the line between AI assistance and human authorship should be drawn and what the technology could mean for the future of writing.
“Current AI-detection tools are not reliable enough to establish authorship”
AI-detection tools should not be treated as definitive evidence of authorship, according to Professor Dhaval Thakker, Head of the Centre for Responsible AI at the University of Hull and Theme Accountability Director for the National Edge AI Hub.
As concerns grow over the use of artificial intelligence in writing, publishing and academia, questions are increasingly being raised about whether existing detection systems can reliably distinguish between human and AI-generated work.
Professor Thakker said the technology remains too uncertain to establish authorship, warning that detector scores can be misleading and should only be treated as an initial signal.
“Current AI-detection tools are not reliable enough to establish authorship, and a detector score should never be treated as proof that text was AI-generated. These systems infer authorship from statistical patterns in language rather than detecting an identifiable ‘AI fingerprint’.
“False positives can occur when writing is highly structured, predictable, grammatically polished or stylistically uniform. Short texts are particularly difficult to classify reliably.”
This could leave some groups more exposed to being wrongly accused of using AI, including students, non-native English speakers and writers whose work has undergone extensive editing.
Professor Thakker said disputes over authorship should instead be assessed through evidence showing how a piece of work was created.
“Useful evidence could include earlier drafts, document version histories, handwritten or digital notes, research materials, tracked changes, timestamps, correspondence with editors, and the author's ability to explain and discuss their work and creative decisions”, he siad.
He pointed to the debate surrounding NeurIPS 2026, a leading AI conference, which used an AI detector in its Position Paper Track and allowed desk rejection without appeal at its highest threshold based solely on the detector score.
Thakker said he would be uncomfortable applying such an approach to authors. Instead, he highlighted provenance, evidence showing how a piece of writing developed over time, as a stronger safeguard.
“Detection can raise a question; provenance, corroborating evidence, and human judgement should answer it,” he said.
“Personal interaction cannot be replicated by AI”
Vaseem Khan, award-winning author who served as the Chair of the UK Crime Writers’ Association from 2023 to 2025 believes the publishing industry faces a difficult balancing act as artificial intelligence becomes increasingly capable of producing convincing prose.
“The integrity of publishing and literary competitions must be preserved. The jury is still out on exactly how to do this,” he says, stressing the damage that could be caused when an honest writer is wrongly accused. “Nothing can be worse than seeing an honest author torn down by a false accusation – but I really don’t know who we would blame in such a case.”
He believes the industry faces an inherent contradiction: “We can’t have it both ways. We want the industry to weed out AI ‘cheats’, yet we don’t want the industry to ever make a mistake, even though we know they are relying on tools that are not perfect.”
While he hopes detection technology will eventually become sophisticated enough to eliminate errors, he says, “But, at this point, no one can guarantee this to be the case.”
For Khan, however, the qualities that make human authors distinctive extend far beyond the ability to produce polished prose. He recalls being shown AI-generated writing modelled on his Malabar House novels, beginning with Midnight at Malabar House, his series set in 1950s Bombay.
“The AI version was close but not quite right,” he says. His books draw on his experience of living in India, extensive research and his interpretation of the period following independence. They also contain a defining personal element: humour.
“These are values that, in my opinion, cannot be replaced by AI – AI is particularly bad at humour!” he says. He also points to the relationship between authors and readers as something technology cannot reproduce. “That personal interaction cannot be replicated by AI.”
Khan believes publishers should play a central role in preventing AI-generated books from overwhelming the market, warning of a future in which AI-generated content could potentially make traditional publishing redundant.
Yet he remains confident that readers will continue to value human creativity. “A beautiful hardback by a favourite author is a valued treasure. The emotions such a book evokes cannot be replaced by a soulless AI writing books.”
On disclosure, Khan draws a firm line around AI-generated prose and plotting. “My personal feeling is that any author that uses AI to generate the actual prose of their writing or for plotting should be upfront about doing so.”
He acknowledges that using AI for proofreading or research is more complicated, but ultimately believes authors know when they have crossed an ethical boundary. “It comes down to an author’s personal integrity.”
“Mistakes are inevitable”
Suman Gupta, Professor of Literature and Cultural History at the Open University, has urged caution over the growing use of artificial intelligence detection tools in publishing and academia, warning that current systems cannot conclusively establish whether a piece of writing was produced by a human or an AI model.
Gupta argues that AI detection remains fundamentally probabilistic rather than definitive, making mistakes unavoidable. “I do not think it is currently possible to conclusively determine whether a given text is produced by human writing or through an AI model's processing,” he says. “I doubt whether such detection can be foolproof simply with reference to a given text.”
According to Gupta, existing tools rely on statistical estimates rather than a clear mechanism for establishing authorship. “All the detection tools at present give probabilistic results, based on statistical estimations rather than some clear causal schema. Mistakes (false positives) are inevitable.”
He is also concerned about the speed with which these technologies are being developed and marketed, arguing that commercial pressures can take precedence over rigorous testing and transparency.
“The current drive is to reduce the margins of these rather than to eliminate these,” Gupta says. He adds that there is a tendency for tools to be “wheeled out for public deployment and marketing quickly, before being properly tested, without putting test results to open scrutiny and repeating tests.”
With many detection systems developed by profit-making companies and growing institutional and legal pressure to use them, Gupta believes the risks deserve greater scrutiny. “Since such tools are produced by profit-making firms, and the institutional and public demands (and increasingly legal requirements) for them are strong, market considerations tend to override robust testing.”
He also questions whether existing regulatory safeguards are strong enough. “The regulatory mechanisms that governments and non-profit bodies offer are weak,” he says, warning that regulation can itself be compromised by the involvement of interested parties. “So, there is good reason to be concerned when such tools are promoted and employed for policing texts.”
Rather than rushing to redefine literary practice because of AI, Gupta advocates a more measured approach. “My inclination is to urge caution before pronouncing on how writers (and others who do creative work) should change their practices.”
He argues that literature has never been judged solely on the text itself, despite claims to the contrary. “I do not think publishers, critics, etc. have ever made their determinations simply with reference to given texts — though they have often pretended that they do.”
For Gupta, the rise of AI demands deeper research into both the technology and the wider structures surrounding it. “Things are not necessarily as they might appear to be or are claimed to be,” he concludes. “But then, I am an academic researcher, what else would you expect me to say?”
