Editorial intelligence versus artificial intelligence for literary fiction manuscript development

This paper investigates generative AI (GnAI) as an editorial mediatory for literary fiction. Applying a digital hermeneutics methodology, our experiment tests ChatGPT-3.5's editing capabilities by comparing its work on a short story to the work of three professional human editors – two working in-house for esteemed Australian literary journals and one freelance, judging for an annual anthology. This case addresses a gap in our knowledge of how large-language-model (LLM) technologies may improve – or assist in improving – literary fiction works, in line with industry-standard editorial conventions, and increase the efficiency and productivity of editorial intervention. The outcomes suggest GnAI cannot yet compare or compete with human 'editorial intelligence' – the result of experience and expertise; intuition, iteration, reflection; and, most importantly, author–editor conversations – in the literary sector. We assert that human management is optimal for literary fiction manuscript development, but surmise that GnAI could have potential application in genre fiction – in conjunction with human oversight (to direct the technology). Future iterations of GnAI will likely have fewer limitations. ChatGPT can, however, enable efficiencies for clearly identified works – those with recognisable tropes and formulaic structures – and has potential to support editors and enhance author engagement with reading audiences.