Photo by Claudio Schwarz on Unsplash
For years, I had a pretty simple writing process. I would write something. Refine it. Read it again the next day. Send it to Tricia. Make a few changes. Then publish it.
It was not especially sophisticated, but it worked. I wrote hundreds of posts that way, and at some point I would decide the piece was ready and send it out into the world.
That last part feels harder now.
These days, even when I write something I am happy with, I feel almost compelled to ask ChatGPT what it thinks. Sometimes Claude too. I will paste in a piece that I already like and ask some version of: Does this work? Is there anything missing? Is there a better way to say this?
Often the answer is helpful. Sometimes it catches something I missed, or suggests a phrase I like better. And sometimes I ignore almost everything it tells me.
But I still ask. And I have been wondering why.
Maybe I just like the validation. It is nice to have something immediately tell you that the piece is strong, the idea is interesting and the ending works. Maybe it is fear of missing something. When another opinion is available in seconds, it can almost feel irresponsible not to ask for it. Or maybe something else is happening, and AI is changing my relationship with my own judgment.
I knew how to decide when something was good before ChatGPT existed. I knew how to write, revise and eventually say, That is good enough. Publish it. Now there is always one more opinion available. One more rewrite. One more suggestion. One more way to make a sentence slightly tighter or an argument slightly clearer. And I wonder if, in all of that help, I am getting a little worse at trusting myself.
I still believe what I have written before about AI and writing. Writing has always been assisted. My dad edited my newspaper columns. Tricia still reads these posts. AI can be another helpful voice in that process. The question I am wrestling with now isn’t whether assistance is good or bad. It is what happens when the assistance is always there, always available, and almost impossible to exhaust.
Dean Shareski had me thinking about this again last week with his post on spelling errors. Dean has always been more comfortible than I am with the rough edges of blogging. A spelling mistake isn’t necessarily evidence that the writing failed. Sometimes it is simply evidence that a human being wrote it.
And maybe that is part of what I am losing. AI doesn’t just offer another opinion. It offers the possibility of sanding off almost every rough edge. A sentence can always be tighter. A word can always be better. A typo can always disappear. Once that level of polish is available all the time, it becomes surprisingly difficult to deliberately leave something alone.
That may be one of the more interesting unintended consequences of AI. We spend a lot of time asking whether it will make us better writers, worse writers, faster workers or lazier thinkers. The subtler question is whether it makes us less comfortable deciding that something is already good.
There is always another prompt we can give it.
“Make this stronger.”
“Give me three better endings.”
“What am I missing?”
“Would you publish this?”
Of course, Dean and I have a certain luxury here. We write our blogs because we like to write. They are passion projects, not job requirements. Nobody is grading them, and neither of us is going to lose an opportunity because of an awkward sentence or a spelling mistake.
Students don’t always have that luxury. Their writing is marked. Grammar and structure matter, and the finished product can shape a grade, an application or an opportunity.
If I have written hundreds of blog posts and still find myself reaching for AI reassurance before I hit publish, it is hard to blame a 16-year-old for doing the same thing when the stakes feel much higher.
Maybe part of the AI conversation isn’t only about teaching people when and how to use it. Maybe it is also about creating more places where writing doesn’t have to be perfect, places where people can try out an idea, write an imperfect sentence, make a spelling mistake and simply keep going.
I don’t think the answer is to stop using AI. I find it far too useful for that. It helps me think, challenges ideas, catches mistakes and often makes my work better. The concern isn’t assistance. It is when assistance quietly starts becoming dependence. But I also don’t want to reach the point where I need its permission.
There was something healthy about the old process. Write. Refine. Send to Tricia. Publish.
The new process can sometimes feel more like: write, refine, ChatGPT, Claude, Tricia, second-guess myself, ChatGPT again, publish. I’m not convinced that is progress.
And yes, I can feel the irony here. Before publishing this post about whether AI has become a crutch, I asked ChatGPT what it thought. Here was the prompt:
I am thinking of a short blog post, The AI Crutch? Even when I write now and I am happy with it I feel compelled to still get an opinion from ChatGPT or Claude. Maybe I just want validation. Maybe I am feeling like I can’t do anything alone. I think there is a short post here. I used to be fine to write, refine, share with Tricia and publish. And even though that worked for so long I have trouble doing that, even with this post that is about using AI as a crutch. Can you feel that irony?
