Remove AI writing patterns, one rule at a time

The patterns, most expensive first

1. "Not X, but Y" — "This is not just a tool, it's a system." The two-beat reveal. A human sentence commits to one thing; this shape commits to a correction nobody asked for.

2. "It's not X, it's Y" — "It's not about speed, it's about trust." Same correction, different grammar.

3. "More than X, it's Y" — "More than a tool, it's a partner."

4. Rhetorical question, then a short answer — "The result? Better." The question exists to be answered by the writer.

5. The rule of three — "clear, direct and honest", "fast, reliable and secure". Three items of the same shape, the same length, ending in a comma.

6. Em dashes used as pauses — a sentence carrying two of them is leaning on the punctuation instead of the syntax.

7. Bold that means nothing — four or more bold spans in one paragraph, each one a label for the sentence right after it.

8. Bullets that follow a template — four or more bullets that all start with a bolded label and a colon. Humans do not format like that by accident.

9. Hedging — "arguably", "to some extent", "in a sense", "it depends", "podría decirse". The sentence refuses to say what it means.

10. The assistant sign-off — "I hope this helps", "let me know if you need anything", "feel free to reach out", "espero que te sirva", "no dudes en preguntar".

11. Dead phrases — "it's worth noting", "in today's fast-paced world", "in conclusion", "dive into", "a wide range of", "at the end of the day", "plays a vital role", "cabe destacar", "es importante señalar", "en el mundo actual".

12. Press-release vocabulary — "delve", "tapestry", "leverage", "seamless", "robust", "landscape", "holistic", "synergy", "paradigm", "myriad", "pivotal", "panorama", "sinergia", "disruptivo", "holístico".

Phrases cost more than single words, and the structural rules cost more than both. A word can be legitimate in your sentence. A phrase built to sound generic almost never is.

Remove them in this order

1. Score the text. The number tells you how much work is in front of you.
2. Work the top of the list first. Start with the "not X, but Y" constructions and the dead phrases, then the words. Cutting the expensive hits first moves the score the most.
3. Rewrite as human, then compare. unslop returns the score before and after and the number of hits removed, so you can see whether the fix worked or you just swapped one habit for another.
4. Read the result. The rewrite is an edit. It is not a promise that the meaning survived intact.

Two things that are not in this list

Unslop does not check facts, sources or plagiarism. It works on the writing, not on the truth of the claims. For a text full of invented numbers and sources, a low score means nothing.

Removing the patterns does not make a text yours. It makes a text readable. If the thinking behind it is thin, the reader will still notice, just later and for a different reason.

A note for Spanish drafts

The Spanish pack catches the same habits in Spanish: "cabe destacar", "es importante mencionar", "en conclusión", "sin lugar a dudas", "juega un papel crucial", "espero que te sirva", "no es X, es Y". The rule set is not a translation of the English one, it was written for Spanish first, because that is where the machine habits are hardest to hear.

Questions people ask

Does this work if I write in English? Yes. English and Spanish are both covered, and the French phrase and vocabulary lists are included too.

How do I know which patterns my own text has? Paste it in. Every hit comes back marked in place with the rule that caught it.

Is it free? 10 scores and 10 rewrites a day, up to 5,000 characters, no account and no key.

How long does it take? The surface pass is instant. The deep read, which is what finds the structural tells, takes up to about 40 seconds on dense text.

Run the list on your own text

Read the rules, then paste your draft and see which ones you actually did. Most people find the first three are enough.