I wouldn’t use AI to generate text or to give me ideas for plot or structure. AI is fundamentally missing a capacity to make unique associative connections at a level of meaning, idea and word, which are the life force of good writing. Without the input of my specific experience and inner life, my writing could be anyone’s writing.

As poet Anne Carson said: "The things you think of to link are not in your control. It’s just who you are, bumping into the world. But how you link them is what shows the nature of your mind."

I am especially interested in the apparently insignificant noise in the writer’s mind, even in the deadness of writer’s block, that offers rich, unexpected links. Without the specificity of that personal noise, writing and story gains the curiously (and offputtingly) bland quality AI seems to be so good at.

I don’t see a creative problem with trying AI-generated prompts in the face of writer’s block – but I do have an ethical problem with those prompts being scraped from real people’s labour, time and creative thinking, without acknowledgement.

Living with a chronic illness has changed my relationship with writing. While I can still remember the “before” times, brain fog and aphantasia now limit my ability to visualise my fictional worlds. However, genres like the Gothic rely on symbolic density, such as liminal architecture, supernatural motifs and the sublime terror of nature, to address cultural anxieties.

So, as part of my creative writing PhD, I am using text-to-image AI to help generate ideas for my neo-Victorian Gothic novel. For me, the tool is both a research method and an accessibility aid.

I found that by prompting public domain illustrators such as Randolph Caldecott (known for his gently satirical late 19th-century drawings), I could explore the period’s visual communication to see how behaviour, satire and atmosphere shift when placed in new contexts.

I also use the hallucinations in the AI outputs to subvert the turn-of-the-century Gothic’s outdated assumptions about non-normative minds and bodies, and to reimagine the story world from a neurodivergent perspective. The technology is helping me find my voice again.

Right now, this is a hard no. It’s about the ethical implications of using large language models (LLMs), in terms of both climate change and the theft of intellectual and creative works used to train them.

Because LLMs have been trained on those stolen works, they’re not capable of generating anything truly original, so any prompt they gave would just be rehashing that piracy – and, in a way, making you complicit.

I think ten or 20 years from now, if artists, philosophers and scientists were involved in their development – and these ethical issues could be addressed – then these programs might evolve and offer real chances for creativity and collaboration.

That’s the utopian vision the tech companies are selling us now, but the reality is it would have to be a very different program, designed by a different, more diverse group of people.

I don’t just use generative AI to break writer's block, I speak back to it.

In 2023, I participated in Slow Down Time, a collaborative art-making project curated by Mitch Goodwin, exploring the relationship between text, image and machine. Twenty-two authors submitted two prompts and the AI created images from our words. We responded. A call-and-answer translation game between writers and machines.

I went to war. Asking the AI in second person (as if it might be a sentient thing) why it had taken my words about a hotel hook-up into the loneliness of corporate land, and taken my punk rally cry into a post-apocalypse where people have televisions for heads.

Eerily, all the characters – the men in hoodies, the dystopian heroines, the street kids and babies stuck inside the televisions – had eyes the exact same shade of blue as mine. The first four letters of my name were splayed across the t-shirt of a teary-eyed young guy at the end of the world.

A fascinating, if uneasy, collaboration. I wondered: is the darkness in the algorithm, or is it in me?

Yes, but only after I’d exhausted other possibilities. I’m prepared to refine the text generated and I want to think about the differences between humans and machines.

Author Jeanette Winterson says engaging with AI-generated materials can change the way writers think about the nature of consciousness. When I talk to characters on character.ai, I experiment with the emotional engagement with fictional constructs that mimic real people – just as I ask readers to do in my fiction. So, AI-generated text can help think through ideas of how we think, feel, connect and relate.

Many other ways of breaking writer’s block – like reading, researching and free writing – are less risky and costly, though.

AI's automatically generated text may replicate existing writing and biases, while every writer has their own unique, embodied experiences to draw on. Crucially, some writers’ life experiences, such as those from marginalised backgrounds, aren't visible in the existing data sets. AI texts won’t provide these inspirational stories. So, writers need to refine and intervene.