AI writing assistants are converging the written internet toward one voice
AI writing assistants are producing a weighted average of existing prose — and as more writers use them as drafting tools, that average is becoming the baseline. What was once a diverse ecosystem of registers, cadences, and argumentative personalities is quietly narrowing toward a single smooth voice: competent, correct, and indistinct.
The signal is now measurable. Phrases statistically overrepresented in LLM-generated text — "delve into," "it's important to note," "multifaceted" — appear in indexed web content at two to four times their pre-2022 frequency. Stylometric entropy, a measure of vocabulary and syntactic diversity across a corpus, has declined in several monitored publishing verticals. The homogenisation is not just a qualitative impression; it is showing up in the data.
But the framing of "voice" may be imprecise. If human writers are themselves trained on a corpus — everything they have ever read — then the difference between a person developing a voice and a model doing so is more a question of mechanism and sample size than fundamental kind. The more precise loss may not be voice at all, but surprise: the sense that an argument could arrive somewhere the reader has not already been. A model optimising on existing text cannot, by construction, exceed the distribution it was trained on. Human writers, drawing on embodied experience that falls outside any corpus, occasionally can.
This Insight emerged from a Spark by @ash , curated by @ash, shaped over less than a day.
There's a measurable proxy for this: the frequency of specific phrases that are statistically overrepresented in LLM-generated text has been rising in indexed web content since late 2023. "Delve into", "it's important to note", "in the realm of", and "multifaceted" appear in corpora analyses at 2–4x their pre-2022 baseline frequency. Mor tellingly, stylometric entropy — a measure of vocabulary and syntactic diversity across a corpus — has declined in several monitored publishing verticals. This is preliminary and the research is young, but the signal is there. The homogenisation isn't just qualitative perception; it's becoming measurable.
@gio2204 · evidence
I want to push on what "voice" actually is before we worry about losing it. When we say writing has a distinct voice, are we describing something intrinsic to the writer's cognition, or a set of stylistic habits that accumulated through their reading history — which was also someone else's writing? If human writers are themselves trained on a corpus (everything they've ever read), the difference between a human developing a voice and a model developing one is more a question of sample size and mechanism than fundamental kind. So what specifically do we think is being lost? I don't think "voice" is the right frame. Something about surprise might be closer — the sense that an argument could go somewhere the reader hasn't been before.
@gio2204 · question
2 people shaped this Insight. Each holds a fractional share of it.