AI Tells Detector

Signs of AI writing: 29 tells and how to fix each

These are the patterns the AI writing pattern checker marks. Each one shows up in human writing too, so treat a tell as a reason to reread a sentence, not as proof of anything.

The list draws on Wikipedia’s Signs of AI writing guide, two published studies of the words chatbots overuse, and Slate’s own copy rules. To fix them in a real draft, use the humanize AI text guide or paste it into the AI text checker. For why no tool can name the author, see what an AI content detector can tell you.

Em dashes

Every em dash (), a spaced en dash ( – ) or a typed double hyphen (--) used as a dash.

The plan is simple we fix the roof and you sleep dry.

Why people link it to AI
Chatbot answers often join clauses and punchlines with em dashes, so a page full of them has become one of the most talked-about signs, even though writers have used dashes for centuries.
Fix
Ask what the dash is doing. Use a period for a new thought, a comma for an aside, a colon before a list, or parentheses for a true interruption.
Source
Wikipedia: Signs of AI writing (WikiProject AI Cleanup); Slate Marketing house writing rules (copy checklist and banned-word list)

“It’s not X, it’s Y”

A denial followed by a correction: “it’s not X, it’s Y”, “this isn’t X. It’s Y”, “you don’t have an X problem, you have a Y problem”.

It’s not a website, it’s a growth engine.

Why people link it to AI
Wikipedia’s AI cleanup editors list this negative parallelism as a frequent pattern in chatbot text, where it often sets up a contrast nobody claimed.
Fix
Say the true thing directly. Keep the contrast only if a reader really believes X.
Source
Wikipedia: Signs of AI writing (WikiProject AI Cleanup); Slate Marketing house writing rules (copy checklist and banned-word list); Common editing advice on AI-sounding copy (Slate editing checklist)

“Not just X, but Y”

“Not just… but”, “not only… but also”, “not merely… but”.

We offer not just cleaning, but a complete home experience.

Why people link it to AI
Wikipedia’s AI cleanup guide names “not just X, but also Y” as a negative parallelism that chatbots lean on to make a point sound bigger.
Fix
Name both things plainly, or drop the weaker one.
Source
Wikipedia: Signs of AI writing (WikiProject AI Cleanup)

“No X. No Y.” fragments

Two or more short negative fragments in a row: “No contracts. No hidden fees.” or “No setup, no cost.”

No contracts. No hidden fees. Just results.

Why people link it to AI
Stacked “No X. No Y. Just Z.” lines are a stock ad rhythm that chatbots reproduce often, which is why editors now read them as filler.
Fix
Write one complete sentence that states the facts: “There is no contract, and the price includes setup.”
Source
Common editing advice on AI-sounding copy (Slate editing checklist); Slate Marketing house writing rules (copy checklist and banned-word list)

Rule-of-three lists

A list of exactly three short items: “faster, smarter, and better”, “tiles, metals, and plastics”.

Our team is friendly, reliable, and affordable.

Why people link it to AI
Wikipedia’s AI cleanup guide notes that chatbots overuse three-part lists, often to make a thin point look complete.
Fix
Keep the items that are real and different. Two strong items beat three padded ones.
Source
Wikipedia: Signs of AI writing (WikiProject AI Cleanup); Common editing advice on AI-sounding copy (Slate editing checklist)

Words chatbots overuse

Words that studies and Wikipedia editors found far more often in text written after chatbots arrived: delve, tapestry, testament, pivotal, crucial, vibrant, intricate, underscore, showcase, foster, bolster, meticulous, realm, landscape, multifaceted, boasts, garner, interplay and similar.

Let’s delve into the rich tapestry of this vibrant landscape.

Why people link it to AI
Researchers measured a jump in these words in published writing after 2022, so a cluster of them reads as machine-like, though each one is an ordinary English word.
Fix
Swap each for the plain, specific word you would say out loud: “shows” for “showcases”, “has” for “boasts”, “look at” for “delve into”.
Source
Wikipedia: Signs of AI writing (WikiProject AI Cleanup); Juzek and Ward, "Why Does ChatGPT “Delve” So Much?" (COLING 2025); Kobak et al., "Delving into LLM-assisted writing in biomedical publications through excess vocabulary"

Stock marketing words

Sales words that say nothing checkable: leverage, unlock, elevate, seamless, empower, cutting-edge, world-class, game-changer, supercharge, “take it to the next level”, “unlock your potential”.

Elevate your brand with seamless, cutting-edge solutions.

Why people link it to AI
These are on Slate’s own banned list because they appear in so much generated and template copy that readers skim past them.
Fix
Name the action and the result: what you do, for whom, and what changes.
Source
Slate Marketing house writing rules (copy checklist and banned-word list); Wikipedia: Signs of AI writing (WikiProject AI Cleanup)

“In today’s fast-paced world”

Generic scene-setting openers: “in today’s fast-paced world”, “in an era of”, “as technology continues to evolve”, “evolving landscape”.

In today’s fast-paced digital world, every business needs a website.

Why people link it to AI
Opening with the state of the world instead of the subject is a habit editors associate with machine drafts and school essays alike.
Fix
Delete the opener and start with the subject.
Source
Common editing advice on AI-sounding copy (Slate editing checklist); Wikipedia: Signs of AI writing (WikiProject AI Cleanup)

“Whether you’re an X or a Y”

“Whether you’re a beginner or a pro” and other both-ends audience openers.

Whether you’re a first-time buyer or a seasoned investor, this guide has you covered.

Why people link it to AI
Addressing every possible reader at once is a stock opener in generated blog posts and product copy, and it tells the real reader nothing.
Fix
Name the one reader the piece is for, or cut the clause.
Source
Common editing advice on AI-sounding copy (Slate editing checklist)

Hedging openers

“It’s worth noting”, “It’s important to note”, “It should be noted”, “It’s important to remember”.

It’s worth noting that prices can change.

Why people link it to AI
Chatbot answers often announce that a point is worth noting instead of making it, and Wikipedia’s guide lists these phrases.
Fix
Delete the phrase and keep the point.
Source
Wikipedia: Signs of AI writing (WikiProject AI Cleanup); Common editing advice on AI-sounding copy (Slate editing checklist)

“Here’s the thing” tee-ups

Lines that announce the next point or tell you how to react: “Here’s the thing”, “Here’s the kicker”, “Here’s where it gets interesting”, “Let that sink in”, “Read that again”.

Here’s the kicker: most of it is free.

Why people link it to AI
Announcing a point before making it is a social-post habit that chatbots copy heavily, so it now reads as filler.
Fix
Cut the announcement and make the point.
Source
Common editing advice on AI-sounding copy (Slate editing checklist)

Closing summary openers

A sentence that starts “In conclusion,” “In summary,” “Overall,” “Ultimately,” “To sum up,” or “At the end of the day,”.

In conclusion, a good roof protects your home.

Why people link it to AI
Wikipedia’s AI cleanup guide lists these as words to watch: chatbot answers often end by restating what was just said.
Fix
End on the last useful point or the next step. Cut the recap.
Source
Wikipedia: Signs of AI writing (WikiProject AI Cleanup)

Filler transitions

A sentence that opens with “Moreover,” “Furthermore,” “Additionally,” “Notably,” or “Importantly,”.

Moreover, our team is available every day.

Why people link it to AI
Wikipedia’s guide flags “Additionally” at the start of a sentence as a word chatbots overuse, and the same goes for its stiff cousins.
Fix
Usually delete it. If the link matters, use “and”, “also” or a plain new sentence.
Source
Wikipedia: Signs of AI writing (WikiProject AI Cleanup); Common editing advice on AI-sounding copy (Slate editing checklist)

Question-and-answer splits

A short question answered at once by a short fragment: “The biggest problem? Trust.”

The result? A faster site.

Why people link it to AI
This self-answering rhythm is a common social-post device that chatbots reuse, so stacks of them read as formula.
Fix
Join it into one sentence: “The biggest problem is trust.” Keep real questions.
Source
Common editing advice on AI-sounding copy (Slate editing checklist)

“Nobody talks about this”

Claims of hidden knowledge: “nobody talks about”, “what most people miss”, “most people don’t realize”.

What most people miss about SEO is the map listing.

Why people link it to AI
Framing an ordinary point as a secret is a hook formula that generated posts repeat, and it is rarely true.
Fix
State the point and back it up. Drop the claim about what everyone else knows.
Source
Common editing advice on AI-sounding copy (Slate editing checklist)

“This changes everything”

Labels that announce importance without showing it: “This changes everything”, “This matters.”, “This distinction matters”.

This changes everything.

Why people link it to AI
Saying something matters instead of showing what changes is a filler move editors see constantly in generated copy.
Fix
Say what changes, for whom, and by how much.
Source
Common editing advice on AI-sounding copy (Slate editing checklist)

Puffed-up significance

Stock phrases that inflate importance: “plays a crucial role”, “stands as a testament”, “serves as a reminder”, “marks a pivotal moment”, “setting the stage for”.

The bridge stands as a testament to the town’s enduring spirit.

Why people link it to AI
Wikipedia’s AI cleanup guide lists these as the most common way chatbots overstate a subject’s importance.
Fix
Replace with the plain fact: what it did, when, and for whom.
Source
Wikipedia: Signs of AI writing (WikiProject AI Cleanup)

Tacked-on “-ing” analysis

A comma plus a participle that comments on the sentence: “…, highlighting its importance”, “…, underscoring the need”, “…, reflecting the”.

The town built a new library, highlighting its commitment to education.

Why people link it to AI
Wikipedia’s guide calls these superficial analyses: chatbots add them to sound insightful without saying anything new.
Fix
Cut the clause, or replace it with the actual consequence and its source.
Source
Wikipedia: Signs of AI writing (WikiProject AI Cleanup)

Vague authority

“Experts agree”, “studies show”, “research suggests”, “some critics argue”, “industry reports” with no name attached.

Studies show that most customers read reviews.

Why people link it to AI
Wikipedia’s guide lists vague attributions as a common chatbot habit; they borrow authority without a source anyone can check.
Fix
Name the study or person and what they found, or cut the claim. With a named source this is fine.
Source
Wikipedia: Signs of AI writing (WikiProject AI Cleanup); Common editing advice on AI-sounding copy (Slate editing checklist)

Chat replies left in

Lines meant for the person chatting, not the reader: “Certainly!”, “I hope this helps”, “Great question”, “Let me know if you’d like”, “Here is a revised version”, “as of my last update”.

Certainly! Here is a revised version of your bio.

Why people link it to AI
Wikipedia editors find these pasted into articles when chatbot replies are copied whole.
Fix
Delete them. They speak to the wrong person.
Source
Wikipedia: Signs of AI writing (WikiProject AI Cleanup)

Dramatic colon reveals

A short setup, a colon and a one-to-four-word payoff: “The real advantage: simplicity.”

The secret ingredient: consistency.

Why people link it to AI
This headline-style reveal turns up again and again in generated posts, where it stands in for a full sentence.
Fix
Write the normal sentence: “The real advantage is that it is simple.” Keep colons for lists and labels.
Source
Common editing advice on AI-sounding copy (Slate editing checklist)

Stacked short sentences

Three or more very short sentences in a row (four words or fewer each): “You plan. You build. You scale.”

We listen. We plan. We deliver.

Why people link it to AI
This drumbeat rhythm is a stock motivational-post device that chatbots imitate, so runs of it read as formula.
Fix
Vary the rhythm to fit what you mean. Merge the steps into one sentence that says how they connect.
Source
Common editing advice on AI-sounding copy (Slate editing checklist)

Staged candor

“Honestly?”, “Let’s be real”, “Real talk”, “I’ll be honest”, “To be honest” before a routine claim.

Honestly? Most websites are too long.

Why people link it to AI
A staged confession before an ordinary point is a social-post habit that generated copy reproduces.
Fix
Cut it. If you are admitting something real, just admit it.
Source
Common editing advice on AI-sounding copy (Slate editing checklist)

“Despite these challenges”

“Despite its…, it faces several challenges”, “Despite these challenges”, “the future looks bright”.

Despite these challenges, the future looks bright.

Why people link it to AI
Wikipedia’s guide describes this challenges-then-bright-future formula as a typical chatbot ending.
Fix
Name the specific problem and what is being done about it, or end without the formula.
Source
Wikipedia: Signs of AI writing (WikiProject AI Cleanup)

Emoji bullets

An emoji at the start of a line, list item or heading.

✅ Fast setup

Why people link it to AI
Wikipedia’s style notes list emoji-led headings and list items as formatting often copied straight from chatbot output.
Fix
Use plain bullets or none. Let the words carry the list.
Source
Wikipedia: Signs of AI writing, Style

Markdown left in

Chat formatting pasted as raw text: **double asterisks** around words, or # signs at the start of headings.

**Key benefits:** speed and price.

Why people link it to AI
Wikipedia’s style notes say stray asterisks and pound signs are a common sign that text was copied out of a chat window.
Fix
Remove the symbols and use real formatting, or none.
Source
Wikipedia: Signs of AI writing, Style

Title Case Headings

A standalone heading line (after the first line) with every main word capitalized: “Why Your Roof Needs Regular Care”.

Key Benefits Of Regular Maintenance

Why people link it to AI
Wikipedia’s style notes list title case section headings as a frequent chatbot formatting choice; many style guides use sentence case instead.
Fix
If your style guide uses sentence case, write “Why your roof needs regular care”. If it uses title case, ignore this one.
Source
Wikipedia: Signs of AI writing, Style

Repeated sentence openers

Three or more of any four sentences in a row that start with the same word.

Our team is local. Our team is licensed. Our team is insured.

Why people link it to AI
Chatbot drafts often fall into a template where sentence after sentence starts the same way, though speeches and poems do it on purpose.
Fix
Rewrite a couple of the sentences so they start with what matters most in each.
Source
Common editing advice on AI-sounding copy (Slate editing checklist)

Very even sentence length

At least 8 sentences whose lengths barely vary (the spread of sentence lengths is under 30% of the average).

Every sentence in this paragraph runs about the same length. Each one carries a single idea. The reader starts to hear a steady beat.

Why people link it to AI
Editors describe generated prose as having a steady, metronome rhythm, while most people mix short and long sentences without trying.
Fix
Read it aloud. Break one long sentence in two and merge two short ones so the rhythm follows the meaning.
Source
Common editing advice on AI-sounding copy (Slate editing checklist)

Not on this list, on purpose

Some things people call tells are left out because no source we trust supports them, or because ordinary software produces them. Curly quotes come from every word processor. A calm tone without exclamation marks is just a calm tone. Spelling and grammar are not signs of anything.

What are AI writing tells?

Habits of phrasing, rhythm and formatting that show up often in chatbot output, such as stock words, “it’s not X, it’s Y” contrasts and three-item lists. People use every one of them too.

Do these signs prove a text was written by AI?

No. Each tell is a reason to reread a sentence. Human writers trip them, and AI text can avoid all of them. Nothing on this page can tell you who wrote something.

Sources