What is the most recognizable ChatGPT tic in English?

Without hesitation: the meta opener "In this article, we'll explore / discover / cover". ChatGPT places it almost systematically on the first line, and it has become the most immediately identifiable marker of an unedited AI text. Removing it alone does a lot for your credibility.

Why does ChatGPT overuse em dashes?

Because it was trained on enormous amounts of English text where the em dash is a common punctuation mark for asides, and it reaches for it by reflex. The issue isn't the dash itself, which is perfectly valid, but the density: ChatGPT scatters one into nearly every sentence, where a comma, parentheses or a semicolon would do the job. That density outside any literary context stays one of the most visible markers of AI writing.

Should you ban bullet lists entirely in a humanized text?

No. A bullet list is useful when you're enumerating genuinely parallel items (benefits, steps, criteria). The problem is the overuse: ChatGPT puts everything in a list, even when a flowing paragraph would read more naturally. The rule: one list per section at most, with items that truly justify the formatting.

How do you remove hollow superlatives from an AI text?

Typical AI superlatives ("strategic", "innovative", "one of a kind", "revolutionary") add volume with no information. The manual method: run a Ctrl+F on these words, and either cut them outright or replace them with a precise fact ("used by 80% of marketing teams" rather than "strategic"). Rephrase does this cleanup automatically with the Pro and Web presets.

8 ChatGPT tics in English (and how to avoid them)

The phrases, turns of phrase and typographic choices that instantly give away a text generated by ChatGPT in English.

There are sentences ChatGPT writes dozens of times a day. In English the repertoire is fairly small: eight patterns are enough to flag a text as AI, for an attentive reader as much as for a detector. Here is which ones, and how to replace them.

1. "It's important to note that…"

Probably the most widespread tic. ChatGPT uses this formula to slip in a detail or a caveat. The problem: no human writes like that, except to sound clever.

Replacement: lead straight with the fact. "It's important to note that the law only applies to companies with more than 50 employees" becomes "The law only applies to companies with more than 50 employees."

2. "In conclusion"

Humans don't announce when they're wrapping up. They wrap up. ChatGPT signals the closing like a teacher at the end of a chapter.

Replacement: cut it outright. Or swap it for a sharp adverb: "ultimately", "in short", "so" (used sparingly).

3. The mechanical "not only X, but also Y"

This structure is grammatically correct, so ChatGPT loves it. But it's heavy and far too symmetrical.

Replacement: a short sentence. "This product is not only effective, but also affordable" becomes "Effective. Affordable too."

4. "Crucial", "pivotal", "essential", "seamless"

These four adjectives are LLM favorites for adding weight. They carry no information. A human uses them once a month, ChatGPT three times per paragraph.

Replacement: either cut them or replace them with a concrete fact. "This product is essential" becomes "This product sells 30,000 units a year."

5. Heavy transitions

"Let's now move on to…", "Moreover…", "Furthermore…", "In this regard, it's worth highlighting that…". All these transition connectors are rarely used by a writer who is at ease.

Replacement: a simple line break, a "still", or nothing.

6. Em dash overuse

The em dash (—) exists in English and is perfectly valid, but ChatGPT scatters it everywhere in business copy, where a comma would do the job. The sheer density is the tell, not the dash itself.

Replacement: comma, period, semicolon, depending on context. One or two em dashes per page reads as a stylistic choice; one per sentence reads as a machine.

7. Systematic bullet points

Faced with any enumeration, ChatGPT pulls out bullet points. Handy in a slide deck, heavy in prose.

Replacement: flowing prose with connectors. "Here are the benefits: speed, simplicity, cost" can become "It's fast, simple and cheap."

8. The "In this article, we'll explore…" opener

Or its variant: "This article explores/presents/breaks down…". ChatGPT was trained on thousands of blogs that open this way. The blogs that actually perform open with a hook.

Replacement: start with the fact that will surprise the reader, not with an outline of what's coming.

The special case of "delve into" and friends

Beyond formula tics, ChatGPT leans on a handful of words that have become AI signatures on their own: "delve into", "navigating the landscape", "in today's world", "tapestry", "testament to". A human writes these once in a while; ChatGPT reaches for them by reflex.

When you spot one, ask whether it carries any real meaning. "Let's delve into the data" is just "Let's look at the data". "In today's fast-paced world" can be cut entirely, because the sentence after it almost always stands on its own.

How to fix it all in one click

Rephrase applies these eight corrections automatically, plus the spacing and punctuation cleanup that keeps your text publishable (em dash density, repeated connectors, hollow superlatives). You paste the text, you pick a preset, and you get a version stripped of LLM tics.

To go further, create a Custom Voice from three of your own texts. The system extracts your signature (favorite connectors, sentence rhythm, register) and applies it to every rewrite. The result sounds like you, not like a tool.

Why it matters

The average reader doesn't catch each tic individually. But they feel the accumulation. On a 1,500-word article that contains four or five of these patterns, they drop off without knowing why. On an online store that publishes 200 product pages all built on the same template, Google ends up devaluing the whole site.

Avoiding these tics is therefore both a question of human readability and an SEO stake. Rephrase automates the first pass; the finishing touch remains your added value.