Common AI writing patterns
Each entry below names a pattern the diagnostic flags, explains why it matters, shows what it looks like in practice, and gives you a concrete rewrite. This is a reference for reviewers and writers, not an evasion playbook.
How to use this library
When the diagnostic flags a pattern, find it in the text before you react to the score. A single flag in isolation proves very little. Flags from multiple families converging on the same passage are what makes a result worth investigating.
AI-scented vocabulary (BANNED_WORDS)
Why flagged: words like "delve," "utilize," "pivotal," "landscape," "moreover," and "comprehensive" appear so frequently in AI output that they become a statistical signal on their own. One instance is nothing. A cluster of them in a short passage starts to look synthetic.
What to verify: is the word doing real work, or is it standing in for a more specific choice the writer never made?
Example — before: "This comprehensive guide delves into the pivotal role that effective communication plays in today's dynamic landscape."
Example — after: "Clear communication reduces misunderstandings. Here is how to get better at it."
Rewrite rule: replace the ceremonial word with the concrete thing. If removing the word loses nothing, it was not earning its place.
AI-typical phrases (BANNED_PHRASES)
Why flagged: templates like "it is important to note that," "in today's rapidly evolving landscape," and "a comprehensive overview of" let a model keep generating tokens before it commits to a specific claim. They are filler disguised as gravitas.
What to verify: does the phrase carry information, or does it just announce that information is coming?
Example — before: "It is important to note that the landscape of AI detection is rapidly evolving, and staying informed is crucial for anyone involved in content creation."
Example — after: "Detection methods change as models change. If you review content regularly, check the method changelog for updates."
Rewrite rule: delete the template. Start with the fact, example, or opinion. If the sentence has nothing left after you remove the template, the sentence had nothing to say.
Throat-clearing intro (META_INTRO)
Why flagged: AI drafts often open by describing what the text will do — "In this article, we will explore..." — instead of doing it. The reader is told to expect value rather than given it.
What to verify: can you delete the first paragraph without the reader losing anything? If yes, the intro is throat-clearing.
Example — before: "In this comprehensive guide, we will explore the various ways in which AI writing detection tools analyze text to determine whether it was generated by an artificial intelligence system."
Example — after: "AI detectors measure word choice, sentence rhythm, and structural patterns, then compress those signals into a score. Here is what that score can and cannot tell you."
Rewrite rule: lead with the strongest concrete statement. Add orientation only if the reader genuinely needs it to follow what comes next.
Formulaic conclusion (META_OUTRO)
Why flagged: endings like "In conclusion, it is clear that..." or "To sum up, the key takeaways are..." restate what the reader just read in ceremonial language. They close the text without closing the argument.
What to verify: does the final paragraph add a new implication, decision, or edge — or does it just echo earlier paragraphs in summary form?
Example — before: "In conclusion, AI detection is a rapidly evolving field, and it is essential for content creators to stay informed about the latest developments to ensure their work meets the highest standards of quality and authenticity."
Example — after: "The tools will keep changing. The question worth asking stays the same: does the text say something a person would actually choose to say?"
Rewrite rule: end with the sharpest remaining point or the implication the reader should walk away with. If your conclusion could be generated by prepending "In conclusion" to your introduction, cut it.
Metronomic rhythm (UNIFORM_RHYTHM)
Why flagged: when every sentence in a passage is roughly the same length — say, 14 to 18 words, paragraph after paragraph — the writing develops a treadmill cadence. Human writers naturally vary: a long setup sentence, then a short punch, then a mid-length elaboration. AI models tend to converge on a comfortable average and stay there.
What to verify: read the passage out loud. If you can predict when each sentence will end before you reach the period, the rhythm is too flat.
Example — before: "AI detection tools analyze text patterns to identify machine-generated content. These tools examine sentence structure and word frequency distributions. The results provide useful signals about the likelihood of AI involvement. Reviewers should interpret these signals alongside contextual information."
(Four sentences: 10, 9, 12, 10 words. Almost identical weight.)
Example — after: "Detection tools look for patterns — word frequency, sentence structure, rhythmic consistency. The output is a signal, not a verdict. Whether that signal means anything depends on what you know about the text, the writer, and the context it was written in, which is exactly why the diagnostic shows its work instead of hiding behind a percentage."
(Three sentences: 11, 8, 35 words. Varied rhythm.)
Rewrite rule: vary sentence length where it improves emphasis, pacing, or clarity. Do not randomize for the sake of randomizing — that is just a different kind of artifice.
Over-signposting (OVER_SIGNPOST)
Why flagged: "First... Second... Third... Furthermore... Finally..." in a predictable ladder signals that structure was imposed from outside rather than grown from the argument. Some signposting is fine. A rigid sequence of it, every few sentences, is a pattern detectors learn to recognize.
What to verify: is the logical order already obvious from the paragraph flow or topic sentences? If so, the signposts are scaffolding that was never removed.
Example — before: "First, you should collect a text sample. Second, you should run the diagnostic. Third, you should review the flagged patterns. Finally, you should compare the results against context."
Example — after: "Collect a sample, run the diagnostic, review the flags. Then compare what you found against what you know about the text."
Rewrite rule: keep signposts that genuinely reduce confusion. Cut the ones that narrate a sequence the reader can already see.
Pivot crutch (PIVOT_CRUTCH)
Why flagged: the "not just X, but Y" construction appears constantly in synthetic marketing and explainer copy. Once is fine. Three times in two paragraphs looks generated.
What to verify: does the contrast actually change the meaning, or is it just adding rhetorical motion?
Example — before: "WROITER is not just a detection tool, but a comprehensive diagnostic platform that doesn't just identify AI patterns, but empowers writers to understand and address them."
Example — after: "WROITER shows you what the diagnostic flagged and why. You decide what to do about it."
Rewrite rule: state the main point directly. Use contrast only when the difference between X and Y is genuinely surprising or non-obvious.
Interpretation
More flags, or stronger repeated flags, mean the text overlaps more with these common AI writing patterns. That is useful diagnostic context — not authorship proof. If you are using this library for policy or moderation decisions, read Limitations first and calibrate against Do AI Detectors Work?.
Calibration links
Cross-check your interpretation: Do AI Detectors Work? for the reliability case, False Positive Hall of Fame for real-world false positives, and How to Spot AI Writing for the reader-side version of these patterns.