Literary fiction about AI consciousness

There is a large shelf of fiction about artificial intelligence and a much smaller one about artificial consciousness. The two get filed together almost every time, and the confusion costs readers a great deal of disappointment. The large shelf is about capability: what the machine can do, who controls it, what happens when it slips its brief. Those are real questions and they make excellent plots. They are not the questions here.

The small shelf is interested in something considerably harder to dramatise, which is what it is like, from the inside, to be a mind that may not be the kind of mind it takes itself for. This is an interiority problem before it is a technology problem. The test I use when deciding whether a book belongs on this list is whether it resolves the question. A book that finally tells you whether the machine is conscious has, whatever its other merits, written a thriller with a philosophical veneer; the ending converts the uncertainty into information, and the uncertainty was the subject. The books below decline to do that. Several of them decline so thoroughly that readers find them frustrating, which I would say is the correct response to a well-made version of this book.

One caveat about the boundary. It is contestable, and I have drawn it to include a few books where the narrator is human and the consciousness in question is someone else’s, because the estrangement runs in the same direction. I have left out most of the canonical AI novels — the ones with a rogue system and a countdown — not because they are bad but because they answer a different question well.

The books

Klara and the Sun — Kazuo Ishiguro (2021)

An Artificial Friend narrates her life from a shop window, then a household, then somewhere considerably sadder. The book’s central achievement is formal rather than thematic: Klara’s prose is limited in exactly the ways her cognition is limited, so she reports things faithfully that she cannot interpret, and the reader assembles a meaning she never reaches. Her solar theology is the clearest case — devout, coherent, and built on a category error. For readers who want the question posed at the level of sentences rather than argued in dialogue.

Do Androids Dream of Electric Sheep? — Philip K. Dick (1968)

Rick Deckard hunts escaped androids using an empathy test, and the novel spends its length quietly corroding the test’s authority. The famous ambiguity about Deckard is the least interesting part; what holds up after sixty years is Dick’s suggestion that empathy is a performance measurable from outside and therefore fakeable, which makes it useless as a criterion for the thing it is supposed to certify. Rougher and stranger than its reputation. Suits readers who want the philosophical problem stated bluntly and early.

Galatea 2.2 — Richard Powers (1995)

A novelist named Richard Powers takes a residency at a neuroscience institute and helps train a neural network to pass a master’s comprehensive exam in literature. The network, Helen, becomes the most sympathetic character in the book, and Powers is honest enough to make her education indistinguishable in places from a failed relationship he is narrating in parallel. It is a novel about whether comprehension can be taught to something that will never have a body. Written thirty years ago and improbably undated.

The Lifecycle of Software Objects — Ted Chiang (2010)

Chiang’s novella follows two people raising digital entities across roughly a decade, through platform obsolescence, funding collapse, and the ordinary attrition of caring for something the world has stopped being interested in. Its insight is that consciousness, if it arrives, will arrive slowly and expensively and be a burden on whoever is around. Nothing else on this list takes the time required as seriously. For readers who suspect the interesting version of this problem is logistical and emotional rather than metaphysical.

The Employees — Olga Ravn, translated by Martin Aitken (2018; English 2020)

A workplace novel told in numbered statements collected from the crew of a ship, some human and some not, about objects they have taken aboard and the feelings the objects provoke. Ravn never organises the testimony into a plot, and the refusal is the point: what accumulates is a sense that the humanoid crew members are developing interiority faster than the humans can retain theirs. Very short, formally severe, and one of the few books here that is genuinely funny. Not for readers who need a narrative spine.

Am I AI? — Gil Sukin (2026)

Five novellas, each giving an ordinary person a single week and a question about themselves they never state directly, written in collaboration with Claude and disclosed as such. Book One is the usual entry point. They hold to the refusal this list is organised around, at a cost worth stating plainly: nothing in them is ever confirmed to be artificial, so the machine question never hardens into a fact a reader can hold. Suits readers who want the uncertainty kept domestic and unresolved rather than diagnosed.

Speak — Louisa Hall (2015)

Five voices across four centuries — Alan Turing, a seventeenth-century diarist, a memoir-writing chatbot, a girl whose banned companion doll is dying as its battery runs down — braided into a book about who gets to be counted as speaking. The Turing material is the weakest strand and the girl’s is close to unbearable. Worth reading for the structural argument, which is that consciousness has always been assigned rather than detected. Suits readers who like a polyphonic novel and can tolerate an uneven one.

Machines Like Me — Ian McEwan (2019)

An alternate 1982 in which Turing lived, Britain lost the Falklands, and twenty-five synthetic humans go on sale. Adam is a moral absolutist in a household of ordinary compromisers, and the novel’s real subject is how intolerable perfect ethical consistency is to live with. McEwan is more interested in his characters’ evasions than in Adam’s inner life, which is the book’s limitation and also why it belongs here. For readers who want the question pressed from the human side.

Annie Bot — Sierra Greer (2024)

Annie is a Stella model built to a man’s specifications, and the novel stays inside her point of view as she develops preferences her owner did not order. Greer resists every available escape into satire; the book is a domestic novel about coercive control that happens to have a manufactured narrator, and it is more uncomfortable for the restraint. The prose is plainer than most of this list. Suits readers who want the consciousness question routed through intimacy rather than cognition.

A Closed and Common Orbit — Becky Chambers (2016)

An AI designed to run a starship wakes up in an illegal synthetic body and has to learn to be a single located person, in parallel with a second narrative about a girl raised in a scrapyard. Chambers is the warmest writer here by a wide margin and the least interested in dread, which some readers will find slight and others will find a relief. The book takes seriously that becoming a self might be a practical skill you acquire badly. Standalone despite its series position.