Trying to maintain a personal or 'authentic' style in the age of AI-generated text can feel like an impossible bind.

For me personally, the em-dash was a source of pride.The only people you saw using it were real readers -- because that's where they picked it up. Em dashes took extra effort to type, but you also needed the sensitivity to language to want to use them at all. Same with the less common but crucial en-dash for ranges. You used it because you didn't want to settle for the incorrect hyphen.

In technical documentation, it was things like carefully back-ticking of file and method names, making tables for things that made sense in tables, bolding and italicizing judiciously for emphasis.

Now all of those, which used to be signs of the care taken in composing a text, are instead read as signs of thoughtless AI generation. Paradoxically, people who put less thought into the form of their written text are more able to continue writing as usual without falling under suspicion.

There's one place where my personal style has managed to remain intact: a preference for longer, more complexly structured sentences. LLMs hate and will always advise against them, favoring maximum punchiness and clarity. But I won't hold my breath that there won't be a reactive stylistic correction in later models that swallows up my long sentences too.

Is anyone else making adjustments to their style? Is the desire to even have a style an obsolete, 20th-century concern? Is it more like 19th-century?