In Defense of Badly Written Emails

(Yes, this was written without AI.)

A not-so-distant time ago, sending a long text message actually cost money.

Brevity had a price tag: a few cents per SMS. Over the course of a year, that added up—roughly the cost of a pizza.

These days, even the cheapest margherita—yes, the one from the place that used to be Egyptian, or maybe still is—runs about twice as much. Inflation has spared no one and nothing, except, somehow, the value of concision.

In the world of written communication, we now live in a state of endless surplus: more words, more sentences, more everything. And all of it free. No cognitive sweat required.

A delight for writers; a torment for readers.

The emails I receive have changed over the past year. First, visually.

They’ve become elongated blocks of text—stacked in three- or four-line bands, speckled with boldface as if marked by a mildly enthusiastic highlighter.

There’s also been a curious evolution in punctuation. The em dash (—) has begun to appear like an introduced species—British, vaguely aristocratic—spreading through my inbox and displacing commas, parentheses, and the humble aside.

I still have no idea how to produce it on my Italian keyboard.

I’ve developed various coping mechanisms for this new textual abundance: scanning only the bold parts, reading one word out of every three, skipping lines, reading only the first and last paragraph, or forwarding the whole thing to a colleague with a silent prayer.

Recently, faced with an especially long document—a PDF report—I gave up entirely and asked ChatGPT to summarize it for me.

The summary was excellent. But as I read it, I caught myself wondering: what if the original author had already run their draft through AI to make it longer, smoother, “better”? What if this was a double-processed text?

So I imagined a kind of accordion-shaped information chain.

A human has something to say and types a quick, messy prompt into an AI model. Raw, unfiltered, straight from the mind.

The AI gets to work. It reorganizes, pads, dilutes, and supplies enough adjectives to fill a small decorative pillow. The text expands and is sent.

What began as a compact, content-dense kernel becomes a wordy little creature. The AI hasn’t added information—it’s simply stretched what it was given. Like a piston opening to fill itself with air.

The recipient, if genuinely interested, then condenses it back down to the essential three or four points. The broth reduces to a handful of atoms. The piston closes, expelling only air.

Writing, though, is not merely the transfer of information. Writing is choosing, arranging, ranking. Writing is a political act (tone, vocabulary) and sometimes an emotional one (occasionally just an exclamation mark).

Writing is the process of taking ideas that float in the air and making them yours—internalizing them, wrestling with them, placing them somewhere meaningful in your mind.

In the AI era, all of this happens at the start, in the prompt. A well-crafted prompt is already a finished piece of writing (which raises the question: why use AI at all?). A sloppy prompt can only yield confusion.

More importantly: outsourcing the work of hierarchy, order, and word choice means never truly owning what you write.

The ideas remain synthetic particles suspended in air. They never lodge themselves in memory or emotion. They are fake—not in the fraudulent sense, but in the sense that they belong to no one.

“ChatGPT, write a note telling my girlfriend I love her” sounds outrageous.

In matters of affection, we prefer shaky grammar and sincerity over polished artificiality.

The flawed, ordinary clumsiness of our own words becomes enchanting when tethered to genuine feeling. This is true in love, and in any context where trust matters—in other words, in all meaningful communication between human beings.

In a future crowded with perfectly structured prose, if you care about me, do me a favor: write to me badly.