The Tells That Make Writing Sound Like AI

People can tell. Not always consciously — often it registers as "this is fine but I do not care" — but the recognition is getting faster, and the cost is real. A reader who spots the pattern discounts everything after it.

The tells are specific and fixable.

Stock openings

"In today's fast-paced world." "In the ever-evolving landscape of." "Whether you're a seasoned professional or just starting out."

These say nothing. They exist because the model is warming up, the way a nervous speaker clears their throat. Cut the first paragraph of almost any AI draft and the piece improves.

Instead: start with the specific thing you actually want to say.

Everything in threes

Three benefits. Three challenges. Three reasons. Three, endlessly, whether or not there are three of anything.

Real thinking is lumpier. Sometimes there are two things that matter and one that nearly does. Sometimes there are seven and you should pick the best four.

Instead: count honestly. If you have two, write two.

Paragraphs of identical length

Look at the shape of the page. AI drafts tend to produce blocks of three or four lines, over and over, like a wall of bricks. Human writing has a rhythm — a long paragraph that builds, then a short one that lands.

Some sentences should be short.

Instead: vary it deliberately. Read it aloud and you will hear where it flattens.

The hollow vocabulary

Delve. Leverage. Robust. Seamless. Unlock. Elevate. Game-changer. Navigate the complexities.

None of these are forbidden words. The problem is that they are doing no work — they are gestures at meaning rather than meaning. "Leverage your existing assets" almost always means "use what you have", which is shorter and clearer.

Instead: if a plainer word exists, use it.

The summary that restates the thing

An AI draft usually ends by telling you what it just told you. In a 600-word piece the reader has not forgotten.

Instead: end on the next action, an honest caveat, or the thought you actually want them to leave with.

The deeper fix: give it something to be specific about

Most AI writing sounds generic because the brief was generic. A model given no facts will produce prose about the category of thing you asked for, and category-level prose always reads as hollow.

Feed it the specifics — the actual client, the actual number, the thing that went wrong last Tuesday — and the output changes character entirely. It is not that the model became a better writer. It is that it finally had something to say.

A working method

1. Write the brief properly, including facts only you know.
2. Let it draft.
3. Cut the first paragraph and the last.
4. Read it aloud. Fix wherever you stumble or get bored.
5. Put back one thing that only you would have written.

Step five is the one that matters. It is also the one people skip.