On 25 October 2026, I posted the following on BlueSky:
I’ve been using ChatGPT for a while. Today I gave it some notes for a study I’m thinking of doing and asked it to draft an outline. It gave me some 1300 words back. I redrafted this in my own style and asked it to analyse my style. It came back with a 2000 word analysis which was very impressive. It will henceforth provide writing in what it considers my writing style rather than its own rather bland style. I also gave it my own Word stylesheet which it now uses to format writing. It is quite impressive. Though I appreciate there are others who think I might be dining with the devil.
Marcus Hill (@drmarcushill.bsky.social) promptly replied with
Not so much dining with the devil as using a technology which is provably detrimental to your own ability to think creatively which is built on environmentally disastrous and economically non-viable infrastructure and a stolen corpus of training data.
Unfortunately, Marcus didn’t link to the research he was referring to, so I did a quick search myself and found several articles with a range of studies, one of which was a systematic review of 28 studies (Holzner, Maier and Feuerriegel, 2025). A surface skim of the research did not appear to support the claim that generative AI was indeed “a technology which is provably detrimental to your own ability to think creatively”. Consequently, I asked ChatGPT
Can you produce for me a broad literature review on Generative AI and creativity, in my own writing style along with references and drop it into my Word stylesheet and link directly to articles. Can you also indicate in which countries the research was located.
What follows is my redraft of what ChatGPT produced. Linking directly to original articles would get over the claim that ChatGPT might invent spurious articles and references.
Generative AI and Creativity
This is a 7000 word pdf of a systematic review of literature on Generative AI and Creativity co-produced my myself and ChatGPT.