Clients and students have been asking for my opinion on whether they can or should use artificial intelligence (AI) in their writing and editing work. This is not a question inviting a quick answer! It depends on so many things, some of which I will introduce here.

What is it?

It’s important to clarify what we mean when we’re talking about AI. Put very simply, there’s software (like Scrivener and Microsoft Word), there’s AI (like ChatGPT and Dall-E) and then there’s software or apps that have options for using AI (like Microsoft Word and Hemingway).

We should also differentiate between traditional AI and generative AI (GnAI). There are now many thousands of different models and applications of GnAI. ChatGPT and BERT are text-based models of GnAI that can be used to create new text based on the user’s request, or prompt. Dall-E is a model of GnAI that generates images from the user’s (text) prompt. Jukebox and Wavenet are models of GnAI that create video.

I should also clarify that I am not dismissing or writing off the usefulness of AI. However, there are technical, practical, financial and ethical issues to consider if you are thinking about using GnAI in your writing or editing work.

Corporate writers and content creators

If you’re producing run-of-the-mill corporate communications and documents, such as daily correspondence, media releases, announcements and reports, you may find it helpful to generate new content using GnAI tools. Check that your employer or client permits you to use such tools in the course of your work for them.

If you’re writing content from scratch (i.e. without AI), then at the basic level you might decide to turn on autocorrect and/or “editing” functions in Microsoft Word. When you’ve completed your draft, add-in tools such as PerfectIt and Hemingway can be used to check consistency in spelling, punctuation and formatting – make sure you check that the language setting is suitable for your intended audience (e.g. Australian English, Canadian English). You may also be able to specify your preferred style guide (e.g. Australian Style Manual).

As you gain competence with these tools, you might experiment in using them to assist in correcting grammar and styling syntax, such as using active or passive voice, or checking for sentence length and complexity. The reading level scores in some of these applications may not be suitable, so ensure you have a deep knowledge of your audience’s demographics and read the available literature on how these scores work in the software.

Non-fiction writers

If you’re writing long-form manuscripts, such as research reports and books for publication to markets in the academic, professional, education and trade sectors, it’s completely fine to use the autocorrect and other settings in tools such as Microsoft Word, if you wish. Remember, however, to check your own work – these tools may pick up spelling errors but they won’t differentiate homophones, for instance (e.g. foreword and forward). Analysing factual information, such as the social implications of a policy statement, still requires human expertise.

Turning on AI

Many of the tools I’ve mentioned now enable users to turn on their AI functions. Make sure you understand what this means in terms of copyright ownership of the work you’re creating, and check with your employer or publisher that using AI is permitted. Seek advice from a qualified legal practitioner who has the requisite experience. The Australian Society of Authors provides a member advice service, including contract reviews, and Copyright Licensing New Zealand provides a legal service for creators in Aotearoa.

Fiction writers, editors and translators

In early 2024, along with three colleagues, I began an investigation into the use of GnAI in the creation and development of fiction. Early results have been publicised in the industry newsletter Books+Publishing and in The Conversation, and in July 2024 the study was formally written up in the journal AI and Ethics. I recommend you read the journal paper for an exploration of the practical and ethical issues.

From this research and my work with authors and editors of fiction and non-fiction, my own conclusion is that current models of GnAI are not yet suitable for optimal work in developing creative literary works. The problems that I have seen arise have been diabolical – texts with no spelling errors, that are technically and grammatically correct, lack the depth, sentiment, vitality and flair of the human writer. One editor described it as “weirdly lifeless”, and I agree. It’s difficult and time-consuming to even attempt to edit a problem that can only be resolved by rewriting.

When a writer entrusts their work to a professional editor, they are mostly concerned that we should maintain their “voice” – where is the author’s voice in a text produced by an AI tool?

The problem is exacerbated because it becomes one of economics. It takes longer to rewrite a text than to edit it. Which means that – for now – using GnAI to create and develop literary fiction is a false economy.