147 phrases across 5 structural archetypes that flag content as AI-written. Built from the Kompozy brand-safety gate — the same rules that govern our autopilot outputs.
The direct answer
AI-generated content flags as AI through 5 structural patterns: hedge words, tricolons (lists of three), pseudo-philosophical bridges, authority hand-waves, and emotional thinness. The 147 phrases in this index cover the highest-frequency manifestations of each archetype. Banning these phrases at output-time is the single highest-leverage defense against AI-sounding content.
Methodology
This is not a study. It is a curatorial index — an opinionated ranking of AI-tell phrases organized by the structural archetype that produces them. The methodology behind the index:
Source corpus. The phrases were collected from three sources: (a) the banned-words list that powers the brand-safety gate inside the Kompozy product, (b) public LLM output across GPT-4-family, Claude-3-family, and Llama-3-family models, generated from generic content prompts during product development, (c) feedback from operators who reviewed autopilot outputs during the calibration window before flipping to autopilot.
Classification rubric. Each phrase was assigned to one of 5 archetypes based on the cognitive shortcut the model takes when producing it. Archetypes were derived inductively from the corpus, not pre-defined.
Tell strength. Phrases within each archetype are loosely ordered by how strongly they signal AI generation when they appear in isolation. A phrase like "It is worth noting that" at the start of a paragraph is a stronger AI tell than "Often" mid-sentence — both are hedges, but one is more diagnostic.
What this index is NOT. It is not a statistical analysis of AI-detection-tool performance. It is not a claim that any specific phrase guarantees AI authorship. Humans use all of these phrases sometimes. The index identifies the patterns that, when concentrated in a single piece of writing, produce the recognizable "this was written by AI" feeling.
Why this matters
Content that flags as AI underperforms in three measurable ways: lower trust (readers discount the claims), lower share rate (people do not amplify generic-feeling content), and lower SEO ceiling (Google\'s Helpful Content algorithm down-ranks content with low specificity scores). Banning the 5 archetypes is not about hiding that AI was involved — it is about producing content that earns the same trust as the best human writing, regardless of how it was produced.
The Kompozy brand-safety gate applies banned-word filtering at output-time, not just prompt-time. This is the structural defense — banned words are stripped from the final output even if the model emits them, then the model is re-prompted to fill the gap.
The 5 archetypes
Click any archetype to jump to its full phrase list and structural analysis.
LLMs are trained to avoid claim-confidence because confident claims trigger hallucination risk during RLHF. The model learned that hedging is "safer." Humans confident in their position do not hedge — they assert. Hedge words are the single most reliable AI tell because they appear in nearly every model-generated paragraph above 100 words.
The phrases (30)
It is worth noting that
It is important to consider
It might be useful to
One could argue that
In many cases
Generally speaking
To some extent
It is often the case that
In a sense
More or less
It depends
It varies
It largely depends on
Broadly speaking
In general
For the most part
In most situations
It is also worth mentioning
It should be noted
It is generally accepted that
Some might say
There are those who argue
Could potentially
May or may not
Tends to
Often
Sometimes
Frequently
Typically
Usually
The fix
Replace every hedge with a specific instance or a confident assertion. "It is often the case that X" becomes "X happened on the last three projects I shipped." Specificity beats hedging on every dimension: trust, engagement, memorability.
2. Tricolons (rule of three)
Why this archetype lands
LLMs love three-item lists because they pattern-match against rhetorical training data (Lincoln, Churchill, Aaron Sorkin). The model genuinely cannot help itself — given a prompt about anything, it will reach for a three-noun structure. Real writers use tricolons sparingly because lists of three lose specificity at scale.
The phrases (23)
Faster, cheaper, and better
Simple, scalable, and effective
Efficient, reliable, and powerful
Engaging, informative, and valuable
Authentic, genuine, and real
Innovative, creative, and bold
Powerful, intuitive, and elegant
Smart, simple, and seamless
Bold, fresh, and original
Insightful, actionable, and clear
Robust, scalable, and secure
Modern, clean, and minimal
Strategic, tactical, and operational
Vision, mission, and values
People, process, and technology
Plan, do, and review
Engage, educate, and entertain
Awareness, consideration, and decision
Strategy, execution, and measurement
Past, present, and future
Quality, quantity, and consistency
Time, money, and effort
Mind, body, and soul
The fix
Cut the third item. If the third item is needed, the first two were wrong. "Faster, cheaper, and better" becomes "Cheaper without losing quality" — a real claim with a real trade-off. The discipline is to commit to two dimensions per sentence, not three.
3. Pseudo-philosophical bridges
Why this archetype lands
LLMs produce these because the training data is dense in essay-form prose. Real writers either skip transitions entirely (paragraph break = transition) or use specific, content-bearing transitions. Pseudo-bridges signal "I am about to say something" rather than actually saying it.
The phrases (26)
In an increasingly digital world
In today's fast-paced environment
In the modern landscape
In the ever-evolving world of
In this day and age
As we navigate the complexities of
When it comes to
At the end of the day
When all is said and done
In the grand scheme of things
It is no secret that
It goes without saying
Needless to say
That being said
With that in mind
Having said that
On the other hand
It is clear that
It is evident that
Without a doubt
It cannot be denied that
There is no question that
In light of
In view of
In the context of
It is no surprise that
The fix
Delete the bridge entirely. Start the sentence with the content. "In an increasingly digital world, marketing is hard" becomes "Marketing is hard." Nine times out of ten the bridge added zero information. The reader already knows we are in a digital world.
4. Authority hand-waves
Why this archetype lands
LLMs invoke vague authority because real authority claims require citations the model often cannot produce. "Studies have shown" is the AI version of "trust me, bro" — a confidence-projection without underlying evidence. Real writers either cite the specific study or do not invoke study-evidence.
The phrases (24)
Studies have shown
Research suggests
Experts agree
According to data
It is well known
Many believe
It is widely accepted
Industry leaders say
The data is clear
Research consistently shows
Studies indicate
Evidence suggests
It has been proven
According to recent studies
Research has demonstrated
Numerous studies confirm
A wealth of research
Experts in the field
Authorities in the space
Top performers
High-performing teams
Leading companies
The most successful
Best-in-class
The fix
Either cite the specific study by name + year, or skip the authority claim entirely. "Studies have shown that X" becomes "The 2024 Backlinko ranking factors study found X." If you cannot cite, the claim was probably weaker than the hand-wave suggested.
5. Emotional thinness
Why this archetype lands
LLMs simulate emotional language by reaching for clinical descriptors of emotion rather than performing the emotion. The result is technically grammatical but flat — "exciting" and "delightful" and "passionate" carry no specific texture. Real emotional language is concrete (a memory, a sensation, a stake) not labeled.
The phrases (34)
Exciting
Delightful
Passionate
Game-changing
Mind-blowing
Revolutionary
Transformative
Inspiring
Empowering
Cutting-edge
State-of-the-art
Best-in-class
World-class
Top-notch
Top-tier
Next-generation
Next-level
Unparalleled
Unmatched
Phenomenal
Incredible
Amazing
Wonderful
Fantastic
Outstanding
Exceptional
Remarkable
Extraordinary
Truly remarkable
Heartfelt
Authentic
Genuine
Resonates deeply
Speaks volumes
The fix
Replace the emotion label with the cause of the emotion. "It was a game-changing experience" becomes "I rewrote the pricing page after that call and conversion went from 1.2% to 3.4%." Emotion comes from the specific concrete; the label kills the emotion every time.
How to apply this index to your writing
Run a banned-word lint on every draft. Manual: paste your draft into a find-and-replace tool with this index as the search set. Automated: use a brand-safety gate that enforces the list at output-time.
Replace, do not delete. Every banned phrase is a placeholder for missing specificity. When you cut "It is worth noting that," the replacement is a specific instance, not nothing.
Track concentration, not occurrence. One hedge word in a 1,500-word piece is fine. Twelve hedges in 200 words is the AI-feel. The signal is concentration.
Calibrate to your voice. Some industries (academic writing, legal copy) legitimately use hedge words. The Persona Brief overrides the universal list with voice-specific exceptions.
Re-check after model updates. Every major model release ships new tells. Re-audit your output against fresh corpora every 6 months.
The brand-safety gate explained — how the banned-word filter integrates with the broader autopilot quality-gate architecture.
AI Brand Voice cluster — the full topic cluster on making AI sound like you instead of like ChatGPT.
Cite this index
Ameen, M. (2026). The 2026 AI Tells Index: 5 archetypes and 147 phrases that flag content as AI-written. Kompozy. https://kompozy.io/research/ai-tells-index
Updates
This index is updated as new model versions ship and new tells emerge. Subscribe via the Kompozy changelog to be notified of revisions.