Novelist Charles Stross Explains Why He Does Not Use AI in His Writing

Charles Stross, author of the Laundry Files series and other genre fiction, published a post laying out his reasoning for not using AI language models in his creative work. The post responds to a recurring assumption: that a technically literate SF author who writes about AI would naturally adopt AI writing tools.

His position, as the post title makes plain, is a deliberate non-use rather than ignorance or inability. Stross has a background that intersects with software and tech culture, so this is an informed refusal, not a reflexive one.

The implications are worth noting for builders in this space. When a technically sophisticated writer with demonstrated understanding of LLM capabilities opts out, the reason is rarely unfamiliarity. It is more likely a judgment about output quality, voice consistency, legal exposure around training data, or the mismatch between probabilistic text generation and the craft requirements of long-form fiction.

Long-form narrative work has specific demands that current LLMs handle poorly: maintaining consistent character interiority across a novel-length manuscript, avoiding the statistical averaging that flattens prose style, and producing the kind of controlled ambiguity that literary tension requires. These are not problems that prompt engineering solves at the margin.

For engineers building writing-adjacent tools, this is signal worth processing. Adoption by technically literate users who understand the stack but still decline to use it points to capability gaps that marketing cannot paper over. The gap is not awareness. It is fitness for the actual task.

The post sits in a growing category of public, reasoned rejections from domain experts who have evaluated current-generation tools against real professional standards and found the fit poor. That category is worth tracking separately from uninformed skepticism.

Source