Language Patterns
September 7, 2026 · View on GitHub
7. AI Vocabulary Words
Watch words: delve, tapestry, landscape, multifaceted, comprehensive, pivotal, testament, intricate, nuanced, leverage, foster, crucial, moreover, furthermore, realm, robust, facilitate, endeavor, resonate, underscore, embark, myriad, paramount, encompass, holistic, synergy, elucidate, culminate, juxtapose, burgeoning
Fire condition: 3+ watch words appear in one paragraph.
Semantic Risk: MEDIUM Preservation Note: Replacing AI vocabulary words changes the specific connotations carried by words like "nuanced" or "multifaceted"; ensure the substituted terms preserve the intended meaning and register.
Problem: These words appear at dramatically higher rates in AI-generated text than in human writing. Individually they are normal English words; clustered together they are a telltale fingerprint.
Before:
This comprehensive report delves into the multifaceted landscape of renewable energy, leveraging nuanced insights to foster a robust understanding. Moreover, the intricate tapestry of stakeholder interests underscores the pivotal role of policy in this crucial endeavor.
After:
This report examines the many aspects of renewable energy in depth, using careful distinctions to build a sound understanding. The complex mix of stakeholder interests shows how important policy is to this work.
8. Copula Avoidance ("serves as")
Watch words: serves as, acts as, functions as, stands as, operates as, works as, remains as, exists as
Fire condition: 2+ copula-avoidance constructions in the same paragraph, or a single instance where "is" would be shorter and clearer.
Exclusion: "Serves as" is acceptable when the subject has a formally designated role function (e.g., "the committee serves as an advisory board" — institutional role, not decorative usage).
Semantic Risk: MEDIUM Preservation Note: "Serves as" sometimes implies a functional or designated role distinct from mere identity; replacing with "is" may flatten a meaningful distinction about purpose or function.
Problem: AI avoids the simple verb "is" by using unnecessarily complex copula constructions. "The park serves as a gathering place" instead of "The park is a gathering place."
Before:
The library serves as a vital community hub, functioning as both an educational resource and a social gathering space. It also acts as a testament to the city's commitment to public access.
After:
The library is an essential community hub, an educational resource, and a social gathering space. It also shows the city's commitment to public access.
9. Negative Parallelisms
Watch words: not just...but, not merely...but also, not only...but, goes beyond...to, more than just...it is, transcends...to become, serial negation ("It wasn't X. It wasn't Y. It was Z."), "Not a X. Not a Y. A Z."
Fire condition: 2+ "not X but Y" structures in the same document, or a single instance where the positive statement alone would be simpler and equally clear. Also fires on a single serial-negation build-up (two or more consecutive "it wasn't/is not X" sentences resolving into "it was/is Z") — the negations are a runway, not content.
Exclusion: Genuine contrastive clarification correcting a misconception ("the event is not a conference but a workshop — no keynotes, only hands-on sessions") — the negative frame is doing real work here.
Semantic Risk: MEDIUM Preservation Note: The negative framing sometimes carries a genuine contrast or correction; collapsing "not X but Y" to just "Y" may lose the implicit rejection of a common assumption the author intended to address.
Problem: AI uses negative-then-positive constructions to make simple points sound profound. Instead of stating what something is, it first states what it is not.
Before:
This initiative is not just a policy change but a fundamental reimagining of urban planning. It goes beyond mere infrastructure investment to become a statement about the kind of city residents want to live in.
After:
This initiative changes policy and fundamentally rethinks urban planning. Its infrastructure investment expresses the kind of city residents want to live in.
10. Rule of Three Overuse
Fire condition: 3-item lists appear 2+ times in the same document, or a triple-part sentence where the count is arbitrary and another count would be equally valid.
Exclusion: Naturally occurring triads in genuinely three-part processes (past/present/future; input/process/output; beginning/middle/end) are not this pattern.
Burstiness note: When rewriting, use the naturally correct count — one strong point if that is all there is, two contrasting items if that is the real structure, four if there are actually four. Varying list counts across a document signals human authorship.
Problem: AI defaults to triple-item lists and three-part structures when the natural count might be two, four, or one. This creates a rhythm that feels rehearsed rather than spontaneous.
Semantic Risk: LOW
Before:
The program fosters creativity, innovation, and collaboration. Participants gain inspiration, practical skills, and lasting connections. The result is a more dynamic, inclusive, and forward-thinking community.
After:
The program encourages creativity and innovation. It also encourages collaboration. Participants gain inspiration and practical skills, along with lasting connections. The community becomes more dynamic and inclusive. It also becomes more forward-thinking.
11. Elegant Variation (Synonym Cycling)
Fire condition: The same entity referred to by 3+ different names or synonyms within a single paragraph.
Exclusion: Legitimate disambiguation (using "the company" vs. "Microsoft" when distinguishing a parent from a subsidiary, or "the study" vs. "the 2023 Stanford paper" for precision) is not this pattern.
Problem: AI rotates synonyms for the same entity to avoid repeating a word, producing text that feels strangely evasive. A city becomes "the metropolis," then "the urban center," then "the municipality" — all within the same paragraph.
Semantic Risk: LOW
Before:
Tokyo is the most populous city in Japan. The metropolis is known for its blend of tradition and modernity. The urban center attracts millions of tourists each year. The Japanese capital continues to grow.
After:
Tokyo is Japan's capital and most populous city. It is known for its blend of tradition and modernity, attracts millions of tourists each year, and continues to grow.
12. False Ranges
Watch words: from X to Y, ranging from...to, spanning...to, everything from...to, whether...or
Fire condition: A "from X to Y" construction appears where the two poles do not meaningfully bound a spectrum — the range is decorative rather than informative.
Exclusion: Genuine numeric or temporal ranges ("from 10 to 100 employees", "from January to March", "from $5 to $50") are not this pattern.
Problem: AI creates artificial ranges to sound comprehensive, often pairing two extremes that do not meaningfully define a spectrum.
Semantic Risk: MEDIUM Preservation Note: Removing a range construction may drop a genuine scope claim about audience or coverage; verify whether the range was intended to convey actual breadth before eliminating it.
Before:
The festival offers something for everyone, from young children to seasoned professionals, spanning everything from traditional folk music to cutting-edge electronic performances.
After:
The festival offers something for everyone, including young children and seasoned professionals. Performances include traditional folk music and cutting-edge electronic music.
32. Comparison Adverb Overuse ("more" without target)
Watch words: more specific, more concrete, more efficient, more effective, more comprehensive, more robust, more seamless, more meaningful, more strategic, more impactful, more nuanced, more proactive, more sustainable, more scalable
Fire condition: 2+ "more + adjective/adverb" comparative phrases appear in one document without a clear target, baseline, or metric. A single instance can fire when it appears in the same paragraph as other formal AI markers such as "comprehensive", "strategic", "framework", "stakeholder", or "in-depth".
Exclusion: Do not fire when the comparison target is explicit ("more efficient than the old process"), when a before/after metric is provided ("reduced latency from 240 ms to 130 ms"), in ordinary quantifiers ("more than 10 users"), or in fixed phrases such as "more or less".
Semantic Risk: LOW Preservation Note: Comparative language can carry a real improvement claim. Preserve the target, metric, or priority if one exists; otherwise replace the vague comparative with the concrete requirement or result the sentence is trying to name.
Problem: AI text often implies progress by stacking "more X" phrases while never saying compared with what. Human prose either names the baseline ("faster than last quarter's build") or drops the decorative comparison and states the concrete change.
Burstiness note: Do not convert every "more X" phrase into the same alternative. Mix direct adjectives, named metrics, shorter verbs, and one retained comparative when the contrast is real.
Before:
The initiative will enable more specific milestones, more efficient resource allocation, and more comprehensive stakeholder alignment. Moving forward, the team should develop a more strategic framework for more meaningful collaboration.
After:
The initiative will enable more specific milestones, more efficient resource allocation, and more comprehensive stakeholder alignment. The team should develop a more strategic plan for more meaningful collaboration.
Example note: The input promises to enable improvements; it does not promise that those improvements will occur. Keep "will enable" for all three items and "should develop" for the recommendation. The comparisons remain vague because the input gives no baseline.
33. Definitional-Metaphor Equation ("X is the architecture of Z")
Watch words: is the signature of, is the shape of, is the language of, is the currency of, is the architecture of, is the backbone of, is the engine of, is the heartbeat of, is the DNA of, is the cornerstone of, is the lifeblood of
Fire condition: 2+ copula sentences of the form "X is the [abstract noun] of Z" appear in the same document/section without concrete support for the equation. A single instance is an audit hint only: do not rewrite unless the same inflated metaphor pattern recurs.
Exclusion: Literal or technical definitions ("water is the universal solvent"), established idioms and textbook metaphors ("the mitochondria is the powerhouse of the cell"), factual "X is the capital/center of Z" statements, and genuine equivalences backed by concrete support in the same passage are not this pattern.
Disambiguation from #8 (Copula Avoidance): #8 fires on text that avoids "is" ("serves as", "functions as" → rewrite to "is"). #33 is the opposite: the sentence already uses "is", but inflates it into an "is the [abstract noun] of [abstraction]" metaphor-equation to manufacture profundity. Do not conflate them — #8 wants the copula restored, #33 wants the empty metaphor-equation dismantled.
Semantic Risk: MEDIUM Preservation Note: The metaphor sometimes points at a real claim ("trust depends on consistent behavior"); when rewriting, recover and state that underlying claim concretely rather than deleting it, so a genuine point is not lost along with the inflated framing.
Problem: AI manufactures depth by equating one abstraction with another through a borrowed structural noun — "signature", "architecture", "currency". The sentence sounds like an insight but asserts nothing testable; swapping the abstract noun ("is the language of" → "is the currency of") barely changes the meaning, which exposes the equation as decorative rather than substantive.
Before:
Symmetry is the architecture of trust. Cringe is the visible signature of moving along a gradient you chose. Consistency is the currency of every relationship that lasts.
After:
Symmetry underpins trust. Cringe visibly marks movement along a gradient you chose. Every lasting relationship depends on consistency.
Example note: The input does not explain its metaphors. This edit stays close to them instead of supplying a new causal theory.
34. False Agency (Inanimate Actors)
Watch words: the data tells us, the numbers speak, the decision emerged, the culture shifted, the conversation moved toward, the market rewards, the market decides, the complaint becomes, the strategy demands, this unlocks, the moment calls for, the work speaks for itself
Fire condition: 2+ sentences in the same document where an inanimate subject performs a human action (deciding, telling, rewarding, demanding, moving a discussion) and the human actor is recoverable from context but never named. A single instance is an audit hint only.
Exclusion:
- Academic/report citation conventions ("the study finds", "the paper argues", "the data show a correlation") — standard metonymy for the authors.
- Genuine agentless processes ("the temperature rose", "the deadline passed", "the market crashed" as an aggregate event).
- Deliberate literary personification sustained as a device, and established idioms.
- Software/system subjects doing what software actually does ("the linter flags", "the pipeline retries").
Disambiguation from #26/#27 (Passive/Nominalization): passive voice hides the actor by grammar ("the decision was made"); false agency hides the actor by promoting an object to actor ("the decision emerged"). Both leave the same question — who did it — but this pattern fires on active-voice sentences, which passive detectors miss.
Semantic Risk: HIGH Preservation Note: Never invent an actor the source does not support. If the text implies the actor (a team, the author, "you"), name that; if no actor is recoverable, restructure around what is actually known ("someone on the team killed the project" → only if the source says so; otherwise "the project was killed within days" may have to survive as-is). Fabricating agency is a meaning violation, not a style fix.
Problem: AI systematically promotes abstractions into actors — decisions emerge, cultures shift, data tells, markets reward — because it lets the prose sound analytical while never committing to who did what. Human writing tends to name the person, or at least stand somewhere ("we", "you", "the team"). The tell is not one sentence but the accumulation: a paragraph where every event happens with no human in it.
Before:
The feedback loop rewarded speed over quality, so the culture shifted within a quarter. Decisions emerged from Slack threads rather than meetings, and the roadmap bent toward whatever the metrics demanded.
After:
The feedback loop favored speed over quality, so the culture changed within a quarter. Decisions were made in Slack threads rather than meetings, and the roadmap followed the metrics.
Example note: The input does not identify the decision-makers. Passive wording preserves that limit; a full agency repair needs source context.