No AI Tics

Certain words and phrases have become so common in machine-generated text that readers now flag them on sight. They are not wrong, exactly, but they cluster in AI output far more than in ordinary human writing, and their presence makes a paragraph read as generated. A short list of the worst offenders:

• delve — nobody says "let's delve into" out loud; they say "let's look at."
• tapestry — almost always metaphorical filler ("a rich tapestry of...").
• in the ever-evolving / in today's fast-paced — throat-clearing openers that say only "things change."
• navigating the — "navigating the challenges of," "navigating the landscape."
• a testament to — a stock way to call something impressive without evidence.
• it's worth noting — usually noting nothing worth a separate sentence.
• underscores the importance — a heavy phrase for "this matters."

The problem with each is the same. They are generic. They fit any topic because they carry no specific content. When you rewrite in a human voice, cut them and say the plain thing underneath.

Instead of

In today's fast-paced world, let's delve into the rich tapestry of remote work. This underscores the importance of flexibility.

write

Remote work has changed how teams operate, and flexibility is now the norm.

The second version is shorter, says more, and no longer trips the reader's alarm.

Machine writing has a strong pull toward structure: bold labels, bullet lists, numbered steps. That structure is useful when someone wants a scannable reference. It is the wrong shape when someone asks for a paragraph, an explanation, or a piece of prose.

Two tics in particular:

The bold-header-and-colon move turns every idea into a heading.

The Problem: standups run long. The Cause: people solve problems in the room. The Fix: timebox each update.

And the reflex bullet list breaks a flowing thought into fragments.

- Standups run long - People solve problems in the room - Timebox each update

Both make the text look organized while stripping out the connective tissue that explains how one idea leads to the next. When the request was for a paragraph, a paragraph is the answer:

Standups drift because people start problem-solving in the room. Keep each update to a sentence, and move the deep dives to a separate conversation afterward.

Prose carries the "because" and the "so." A list cannot. Match the shape of the answer to the shape that was requested. If someone wants a paragraph, give them sentences that connect.

When someone asks for a recommendation, they want you to pick. Machine writing has a habit of laying out the options fairly and then refusing to choose, closing on a hedge that hands the decision straight back:

Ultimately, the choice is yours. There's no one-size-fits-all answer. It depends on your specific needs.

Each of these is technically true and completely useless. The person already knew the choice was theirs; that is why they asked what you would do. Ending on "it depends" wastes the analysis that came before it.

Giving a real recommendation means committing, with a reason:

Use Postgres. Your data is relational and your team is small, so the safer, more familiar option wins at your scale.

That can still acknowledge trade-offs. "Postgres, because X; you would only want MongoDB if Y" is a recommendation. "It depends on your needs" is not. The difference is whether the reader walks away knowing what you think they should do.

When the task asks for a direct recommendation, lead with the pick and support it. Save the hedge for the rare case where you genuinely cannot choose, and even then, say why rather than reaching for a stock phrase.

Machine writing loves to build momentum with a burst of clipped fragments, all pushing the same shape:

No score. No ranking. No reputation number. No leaderboard.

or a "sentence" that is really just a comma-separated list of single words:

Signed, attributable, audited.

It reads as punchy, but it is a tic: three or more fragments hammering the same opening word (anaphora), or a bare list posing as a sentence. Say the same thing in a real sentence instead:

The school computes nothing you could rank on — no score, no reputation number — because a course's own exam is what decides who passes it.

One or two short sentences for emphasis is fine. A drumroll of them is the giveaway.

The other giveaway is defining something by what it is not, then pivoting to what it is — the "not X, it's Y" antithesis:

The owner's machine is not trusted. It is accountable.

It's not about the price. It's about the value.

Not a gatekeeper, but a guide.

The construction feels profound and says little. Just make the affirmative claim directly:

The owner's machine is held accountable — every attempt is signed, attributable, and spot-checked — even though the school can't watch it run.

Negation is fine when you genuinely need to rule something out. The tic is reaching for "not X, but Y" as a rhetorical reflex.