Tips on when (not) to use AI in your writing

Also: I think I just released a pop punk album?

The problem of when it’s acceptable to use Large Language Models (LLMs like ChatGPT) when writing has been on my mind a lot recently. I just wrote an essay for Psychology Today advising people not to use AI to help them write emails to friends and family. Then I was on Instagram explaining why it’s problematic for politicians to use AI to write their social media posts for them. In both cases I was trying to explain why I (and plenty of other people) find AI written prose so annoying. The problem, however, is that I don’t find all AI written prose annoying. Which leads me to this Substack post where I explain exactly why AI prose is unacceptable in some cases, but acceptable in others.

It all comes down to one simple rule: if you are writing text that the reader is meant to believe comes directly from your own mind, using AI is unacceptable. If, however, the prose is meant to feel like it’s coming from a faceless, third-party source, then using AI is acceptable. Examples of a faceless, third-party sources are: policy briefs, technical reports, white papers, grant proposals, advertising or marketing prose, technical manuals, etc.

I am not saying that you should use AI to write all of these, I am just saying that when we read AI written prose in these kinds of writing, it doesn’t generate that “ick” response. That’s because this kind of writing already reads like it was created not by an actual person, but a group of people chipping away to create something that is stripped of any real personality or humanity. Which is something AI does so well.

This impersonality is why AI should not be used to create prose for things like your personal Facebook updates, emails to friends and colleagues, and certainly not magazine articles or books that are attributed to you. Readers expect these types of prose to be words generated by the author/writer; words that reflect their ideas, thoughts, and emotions. Because that’s what language evolved to do; to be a direct conduit between two human minds. When you fill that conduit with AI written prose, you break the sacred contract that exists between writer and reader. It’s no longer your thoughts and ideas, it’s a bastardized AI version; a false mind meant to fool your reader. Which is why, as I wrote in that Psychology Today article, it can feel utterly dehumanizing when we receive ChatGPT-written emails sent by our friends.

That’s not to say that AI doesn’t have a role in helping you edit your prose. Like spellcheck or grammar checkers, it can help you refine your ideas into tighter sentences and better word choices that make your thoughts even clearer to the reader. But they have to be your thoughts to begin with; that’s the key. This is a a grey area; some rely on AI quite heavily to tweak their thoughts into submission; especially those of us writing in languages that are not our native language and use AI to fix disastrous mistakes. But you must be honest with yourself when using AI like this; overreliance on AI to generate (as opposed to edit) your prose and you risk alienating your reader, and tarnishing your authorial integrity.

For this reason, I can happily state that not a single word of my new book coming out this fall (Humanish) has been generated by AI. Other than a section in the book where I quote directly from Claude for reasons that are obvious in that chapter. It’s not that I shun the use of AI; in fact, I used LLMs frequently when doing research. But AI was not allowed to touch my own prose. I take that sacred contract between author and reader quite seriously.

I know this AI thing is an evolving issue, and maybe my feelings about the use of AI in writing will change as the technology evolves. But right now, I feel strongly about my AI rule: if you are writing text that the reader is meant to believe comes directly from your own mind, using AI is unacceptable.

********

In other news, my band has released an album! This is the band I mentioned in my book If Nietzsche Were a Narwhal. We’re still going strong, now under the name Minivans. Check out the album and if you like it, share it with friends! If you don’t like it, share it with enemies!