Statistical detectors call this low perplexity: model-written text chooses high-probability, unsurprising words and so reads as predictable. Cover letters happen to be the worst-case use case for this — they're a genre where the conventions are so strong that the high-probability next word is almost always the cliché.
The cliché library
- "I am writing to express my interest in…"
- "As a passionate and driven professional…"
- "I bring a unique combination of skills and experience…"
- "I would be honoured to contribute to your continued success…"
Each of those is the statistically likeliest next sentence given the cover-letter genre. A model that has learned the genre will produce them. A hiring manager who has read ten thousand cover letters will recognise them in half a second.
How to write something less predictable
Start with a concrete fact about your own work that the model could not have invented. The number you shipped, the bug you found, the customer you lost and what you learned. Predictability collapses the moment your sentence could only have come from your particular life.
Before
I am writing to express my interest in the Senior Analyst position. As a passionate and driven professional with extensive experience in data analytics, I bring a unique combination of skills that would make me a valuable asset to your team.
After
I spent eighteen months as the only analyst at a Series A logistics company, where most of the job was convincing the founders that the dashboard they wanted would lie to them. I shipped the one they actually needed. That's the kind of work I'm looking for next, and the Senior Analyst role looks closest to it.