Nesrine Malik warns against AI as writing partner

More than half a dozen fake or misattributed quotes made it into a book after an author used an AI system as a “research partner”, according to Nesrine Malik in The Guardian. Malik points to that episode alongside other recent cases in which journalists and writers were tripped up by AI-generated material, including fabricated quotations and texts suspected of being machine-written.

Malik’s argument is not that errors are a temporary glitch on the way to better models, but that the tools are being adopted precisely where the cost of being wrong can be pushed onto someone else. In newsrooms under time pressure, a plausible-sounding line that “checks out later” is a tempting shortcut; when it doesn’t, the correction burden lands on editors, readers, and the broader information environment, not on the vendor selling subscriptions. The same dynamic shows up on social platforms, where Malik describes AI-written posts as already “rampant”: the incentives favour volume and authoritative tone, not provenance.

She also lingers on the aesthetic signature of machine text: bland, repetitive language familiar from customer-service scripts, press releases, and social posts optimised for engagement. That tone fits an online politics increasingly built around rapid reaction and low-effort signalling. Malik argues that writing is not just output but a process—shaped by a person’s history, relationships, and the struggle to find an image or a phrase—and that outsourcing parts of that struggle risks weakening the writer’s own capacity to think.

The piece ties this to a broader cultural shift in which “content” is plentiful while trust is scarce. When AI can imitate existing styles, Malik suggests, it becomes easier to flood channels with passable text than to produce original reporting or a distinctive voice. She cites research indicating that leaning on large language models can reduce mental engagement, a concern that matters less to institutions rewarded for throughput than to individuals trying to keep their judgement sharp.

Malik’s own response is practical: she avoids using AI tools for writing, partly out of fear of importing their cadence into her work, and partly because the risk of subtle errors compounds when machine text is treated as a neutral helper rather than a fallible source.

In Malik’s telling, the technology’s most reliable feature is not speed but deniability: when the words are wrong, there is always a system to blame.