Last modified: 2026-09-26 05:55:30 (UTC)
AI-Assisted Authorship
This chapter was written with the assistance of GitHub Copilot, which expanded on outlined ideas and draft notes provided by the author. The content represents the author’s perspective and has been reviewed for accuracy, but the detailed prose was generated through AI assistance.
AI writing assistants have a recognizable house style. When you let one draft your prose, it leaves fingerprints: words, sentence shapes, and formatting habits that recur across unrelated documents. Readers who have seen enough machine-generated text learn to spot these patterns, and once they do, the writing reads as generic and unconsidered, even when the underlying content is sound.
This chapter catalogs the most common tells so you can remove them. It is written mainly for AI assistants drafting scientific prose, but it is just as useful for humans editing AI-assisted drafts.
One caveat first: no single word or pattern below is wrong on its own. Each has legitimate uses. The signal is clustering and mechanical repetition, the same constructions appearing again and again regardless of what the sentence needs. Edit for the cluster, not the isolated instance, and do not flatten your prose into a voiceless register trying to avoid every word on a list.
Some words appear far more often in machine-generated text than in careful human writing. Many are also Latin-derived words with plainer alternatives, so the advice here overlaps with the Word choice chapter. Table 1 lists frequent offenders and plainer replacements.
| Overused | Plainer alternative |
|---|---|
| delve into | examine, study |
| leverage | use |
| utilize | use |
| showcase | show |
| robust | reliable, well-tested |
| seamless | smooth |
| pivotal, crucial | important, key |
| testament to | shows, demonstrates |
| tapestry, landscape, realm | field, area, or name the thing |
| navigate | handle, work through |
| underscore, highlight | show |
| foster, bolster | encourage, strengthen |
| intricate, meticulous | detailed, careful |
| comprehensive | complete, or say what it covers |
| groundbreaking | new, or say what changed |
| unlock, harness, empower | enable, use, allow |
| myriad, plethora | many, or give the count |
| seamless, seamlessly | smoothly, or say what does not break |
| holistic, multifaceted, nuanced | say which parts or distinctions |
| paramount | most important |
| embark, elevate | start, improve |
| streamline | simplify, speed up |
| synergy | say what the combination does |
| actionable | say what action it supports |
| beacon | example, model |
| game-changer, gamechanger, state-of-the-art, cutting-edge | say what changed and by how much |
| ever-evolving | changing |
| treasure trove | collection, source |
Whole phrases recur too. “In today’s fast-paced world”, “it is important to note that”, and “plays a vital role in” add length without content; delete them and start with the actual point. Two more families recur, and each member should give way to the literal claim:
AI assistants reach for a few sentence shapes by reflex, whether or not the content calls for them.
The strongest tell is the not just X, but Y antithesis. It promises a profound contrast and usually delivers a hollow one.
Example 1 (The “not just X, but Y” reflex)
❌ This method is not just fast — it is transformative.
✅ This method runs in half the time of the previous approach.
The plain version makes a concrete, verifiable claim. The antithesis frame makes none.
Other reflexes to watch for:
Beyond the antithesis, AI drafts lean on a set of sentence shapes that stage a point instead of stating it.
Example 2 (A staged point)
❌ It can read like a definition that says nothing. Its use is in what it forces you to name: a project that cannot say what its \(E\), \(T\) and \(P\) are has not yet stated a machine learning problem.
✅ A machine learning problem is stated only when its experience \(E\), task \(T\) and performance measure \(P\) are named.
The first version packs six tells into two sentences:
The second version states the claim once.
Watch for these shapes:
Four more habits blur the referent or delay the claim. Other chapters cover each:
Some tells are typographic rather than verbal.
**Term:** explanation bullet is fine once, but applying the pattern to every list turns it into a tic. Use it when the label earns emphasis, plain bullets otherwise.The last group of tells is about register.
A short scan for these patterns before you submit a draft catches most of them. Cut the filler and the reflexes, but keep your own voice: the goal is clear, honest prose, not prose that has been sanded smooth.