A reader writes:
I am a writer/editor for a relatively large organization that invests a significant amount of resources in targeted messaging. Obviously the labor market broadly has been rough on writers and editors over the last few years, and I genuinely enjoy my job, my compensation/benefits, and my team.
I am not exaggerating when I tell you each piece of copy I produce has between 5-10 reviewers across departments. My products have to be approved at each stage of the outline, flat draft, design layout, and final print stage. We have editors with PhDs, and we have training budgets to attend conferences and learn from many highly respected word, grammar, and marketing nerds. My business writing and technical editing has improved dramatically over two years here. I have redrafted this email multiple times for tone and clarity. When I find a mistake after pressing send, I WILL cringe.
Like many other workplaces, our organization began encouraging integrating AI across our departments to streamline workflows. The org hosts regular AI training sessions with our IT administrators, which I’ve attended. They are hit and miss in terms of helpfulness. I’ve experimented but never found our office AI to be particularly useful for research and outlining, and I’m pretty skeptical of relying too much on anything for editing beyond spellcheck (which overall seems to have gotten worse!).
The problem with heavily edited writing is that lots of readers now suspect or assume the content is AI-generated. I see this most often on Reddit: anytime I use an em-dash or structure a comment “too well,” I’m accused of being an AI bot. Since spellcheck on my brand new phone is abysmal (why??), I’ve just stopped correcting most of my typo errors on social media as a “real person” signal. Obviously, lowering my standards at work is not an option.
Here’s where I’m paranoid: there’s a lot of talk in the public sphere now about using AI writing detectors on published pieces (like Pope Leo’s encyclical on AI), and a lot of the common advice doled out on social media to spot AI generation is standard pace/grammar/structure in our business writing. I have had another editor ask me if I used AI when I swapped the word “combined” with “unified” in a piece! It seems like only a matter of time before something I produce will be legitimately scrutinized for an AI origin, and if that happens, I think those online writing detectors would call my work AI. But I’m the writer! What do I do if someone accuses me of generating my work?
I don’t think anyone has a good answer to this, and it’s something that nearly all writers are struggling with right now.
It’s absolutely true that people think they can identify AI writing based on things that don’t indicate it at all (like em-dashes).
Until recently I would have said that if you’re a good writer yourself or even just a prolific reader, you can pick up on a sort of AI texture to writing that clearly gives it away — but then I read an article at Serious Eats that I was sure was AI-generated and apparently others thought it too and called it out, and one of their long-time editors, a writer I respect and consider credible, said it absolutely wasn’t (and the author does seem to have a consistent writing style predating AI). So now I’m questioning my own ability to detect it, when previously I was confident I could tell.
And of course, AI trained on human writing, so its “tells” are things we all did in the writing they stole to train it on.
None of which answers your question, but I don’t think this is solvable right now and you shouldn’t let it make you a worse writer meanwhile. The best you can do is to have a clear policy for your organization and your department that you don’t use AI-generated writing and point people toward it if questions are raised. For some people, that will be enough; they want to know it’s something you’ve considered thoughtfully and taken a stance on. For others it might not be, but other than pointing out that AI detectors also concluded the U.S. Constitution was written by AI (in 1787!), I just don’t know that there’s anything you can do about that.