A Week of Writing Without AI (What I Missed, What I Did Not)

The experiment was simple and slightly terrifying: one week, real deadlines, no AI assistance for any writing, drafting, editing, research summarization, nothing. Just me and the blank page like it was 2019. I wanted an honest answer to a question that had been nagging me: had the tools made me a better writer with a faster workflow, or a worse writer with a faster dependency?

The first two days answered the dependency question uncomfortably. My drafting was slower, obviously, but the surprise was the hesitation: I had gotten used to outsourcing the cold start, and facing a blank page without a first-draft machine produced real procrastination I had not felt in a year. The second surprise was a small vocabulary of crutch phrases I kept almost typing, the connective tissue that AI drafts had taught my fingers. Neither discovery says anything bad about the tools. Both say something specific about how I had been using them.

By midweek, the quality picture split by task type. Writing that depended on my judgment, arguments, opinions, anything with a position, came out better unassisted, not because the AI makes prose worse but because drafting is where my thinking happens, and outsourcing the draft had been quietly outsourcing the thought. I had noticed this before at the edges; a week without the crutch made it undeniable. The sentence I struggle to write is the sentence I needed to struggle with.

But the experiment also found real losses. Editing, my least favorite and most delayed task, had genuinely improved with AI: the tool catching my passive constructions, flagging my bloated paragraphs, and shortening my forewords was pure gain, and doing it unassisted took longer and produced worse outcomes by my own editor's judgment. Summarizing long source material for research was the same, slower and worse without help. The tools had not replaced my writing; they had replaced specific chores inside it, and those chores were real.

The week ended with a permanent reconfiguration rather than a verdict. AI stays out of my first drafts of anything with a position, permanently, because the struggle is the work. AI stays in my editing passes, my summaries, and my research digestion, where it is a better assistant than my worse self at midnight. And I added one habit the experiment suggested: drafting by hand first, then using AI as an editor rather than a ghostwriter, which produces work that is recognizably mine at twice the previous speed. The tools did not make me a worse writer. Letting them write the thinking parts did, and that was a choice, which means it is fixable.