The Spark AI Can’t Fake.

A story came over my desk this week that focused my mind into the theme that eventually became this week’s blog post. The Commonwealth Foundation had spent weeks reviewing allegations of AI use in the 2026 Short Story Prize. They spoke in detail to every regional winner, examined working drafts, time-stamped documents, and notes, and consulted the judges. Their conclusion, in their own words:

We held detailed discussions with all regional winners about their creative process, and they collaborated fully in our review. We also examined evidence related to the development of their stories, including working drafts, time-stamped documents and notes. After a thorough consultation with our judges and careful consideration of all available information, we are satisfied that AI was not used to write the winning stories.

The row had centred on Jamir Nazir’s “The Serpent in the Grove,” the Caribbean winner later published by Granta. AI detectors had flagged it, accusations spread, and Granta eventually stepped back from its publishing partnership with the prize over questions of editorial control. The Foundation’s review was thorough and, on the evidence they saw, cleared the winners. But the episode itself felt like a sign of the times: writers who should have been celebrating were instead forced to assemble alibis.

I’m all for authors being transparent about how they work, but I’m not sure investigations of this sort profit anyone in the long run. The direction of travel is unmistakable: AI is becoming an ordinary tool, in much the way a spell-checker became ordinary, or a thesaurus before that. Insisting it must not be used at all is a position with a short shelf life. The judgement should be made on the end product, not on the apparatus that produced it.

The real question is not just whether AI was used. Instead, we should ask: does the finished piece matter? Does it move anyone, disturb anyone, or tell anyone something they did not already know about being alive?

What AI genuinely struggles with is ideation, meaning the originating pressure that makes new artistic forms necessary in the first place. It can remix existing genres at industrial speed. It can produce a passable detective story, a passable fantasy quest, a passable column, or a passable press release. However, it has no childhood, no private grief, and no generational moment pressing on it. It has no reason to need the world to be otherwise. It is an immense, well-punctuated echo chamber that can inhabit nothing because it has never had to live anywhere.

That difference becomes clearest when you look at moments in history when humans invented entirely new artistic languages because the old ones no longer told the truth.

Photography did not kill painting. It simply killed the monopoly of one particular kind of painting: the kind whose highest skill was making a flat surface look exactly like the visible world. Once a machine could fix an image quickly and cheaply, painters were released, or perhaps forced, to ask different questions. What does light feel like, rather than merely look like? What does it mean to see a city, a crowd, a body, or a dream from multiple angles at once? The answers arrived as Impressionism, Post-Impressionism, Cubism, Dada, and Surrealism. Photography did not cause every one of those movements in a straight line, but it fundamentally changed the creative weather. It made one kind of labour less scarce, and artists had to decide what else they were for.

Something similar happened in music in West Germany after the Second World War. A generation coming of age wanted a sound that belonged uniquely to them and to their historical moment. The blues pentatonic scale, the twelve-bar form, the swung rhythms of Anglo-American rock and roll — none of it felt like an honest answer to the question “what should our music be now?” So they deliberately rejected those influences and developed something uniquely their own.

What they built instead relied on repetition as propulsion rather than laziness. They used modal harmony that barely moved, long drones, tape manipulation, early synthesizers, and the motorik beat. That straight, relentless four-to-the-floor pulse sounds less like a drummer swinging around a groove and more like a train that has developed strong opinions about forward motion. Can could be loose, funky, and improvisatory; Neu! locked into a hypnotic forward drive; Kraftwerk moved toward precision electronics and the romance of machines; Popol Vuh wandered into mystical drones. The British music press, baffled by the phenomenon, labelled it Krautrock. It was not a single uniform sound, but rather a family of refusals that ended up rewriting the grammar of everything from techno and ambient to synth-pop and industrial music.

AI can now be prompted to “make something Krautrock-ish” and it will do a respectable job. What it cannot do is wake up in a particular country at a particular moment in history and decide that the available musical languages are lying. That decision still requires people.

We are seeing a similar hunger today. With generative tools flooding culture with competent but often weightless content, audiences are reaching for work that feels lived-in, meaning work made by specific bodies, specific histories, and specific, private logics. That is why the masked Quebec duo Angine de Poitrine has connected so strongly. They are two humans from Saguenay playing microtonal math-rock on custom guitars with twice the usual number of frets per octave. Dressed in polka-dot outfits and oversized papier-mâché heads, they make music that is rhythmically knotty, physically demanding, and gloriously unmarketable. It sounds like nothing an algorithm would have suggested on its own. It sounds like two people who simply needed to make exactly this.

None of this implies that the only honourable response is to swear off AI and chisel everything by hand. In fact, the opposite is true. The sensible approach to any powerful tool is to master it and exploit what it is genuinely good at. AI can accelerate the administrative and tedious parts of writing, such as research, structuring, rough drafts, variations, and proofreading. This allows finite human attention to stay focused on the part only humans can execute: deciding what is worth saying and why it matters right now.

Consequently, publishers, editors, and curators will matter more, not less. When production becomes frictionless, the cultural bottleneck moves to discernment. There will be vastly more material to sort through, and while the surface will often look polished, the wheat will still need to be separated from the chaff. The people whose job it is to find the real gems will probably have to work harder than ever before. That is not a crisis; it is simply the new shape of literary labour.

In the long run, the real story is expansive. We are moving toward a time when technical craft mastery is no longer the main gatekeeper to creativity. More people than ever will be able to give form to what they actually want to say. The danger is not the tools themselves, but rather letting a handful of corporations decide which versions of human experience get to speak through them.

The eventual victory will belong to the generation that grows up treating AI as unremarkable infrastructure, much like electricity or the internet, and spends its energy entirely on the only thing that was ever truly scarce: the lived, felt, awkward, and necessary human reason to make something new in the first place. Our job right now is to ensure the tools amplify that reason instead of flattening it.

That, rather than the purity tests, is the work worth doing.