AI
AIMEVKA
All Articles
creativity

How to Spot AI-Generated Text? It's Not About Em Dashes

7 августа 2026 г.·3 min
How to Spot AI-Generated Text? It's Not About Em Dashes

AI-generated content is getting harder and harder to tell apart from human writing. A recent YouTube video, “What Gives AI-Generated Text Away? Not Em Dashes,” offers a fresh perspective: habits like using long em dashes aren’t the real giveaway anymore. For AI creators and marketers, this is a wake-up call—methods for detecting and disguising AI text are evolving fast. Let’s break down what actually reveals AI writing, why it matters for business, and what practical steps you can take.


Why Distinguishing AI Text from Human Writing Matters


While some debate whether AI writing has a “soul,” others use AI for everything—ads, blogs, newsletters, education. But here’s the catch: audiences are more discerning, and platforms are getting better at detecting auto-generated content.


  • Google can downgrade sites overloaded with “machine” articles.
  • Social networks crack down on bots and spam.
  • Partners and clients value authentic, human communication.

So, if you create content with AI, making your output sound “natural” isn’t optional—it’s essential.


Quick Recap: Em Dashes Aren’t the Smoking Gun


The video debunks a popular myth: supposedly, AI loves em dashes (—). But the creator shows this is outdated thinking. The true “tells” of AI go much deeper.


Why the Em Dash Theory Falls Flat


  • The latest LLMs (GPT, Claude, etc.) use punctuation just like humans do.
  • Em dashes appear in both human- and AI-written texts.
  • Avoiding em dashes won’t magically make your writing seem more human.

Real Signs of AI-Generated Text


Forget old markers like em dashes. Experts point to more reliable clues:


1. Repetitive Patterns


AI can string together readable text, but it often repeats the same structures or stock phrases—think “in conclusion,” “it should be noted,” “it’s important to understand.”


2. Overly Polished Structure


AI text is almost too perfect—no typos, no odd turns of phrase, no bursts of emotion. It’s logical and tidy but lacks human quirks.


3. Lack of Fresh Insights


AI rarely comes up with unexpected thoughts or personal experiences. Expect lots of generic statements, few details, and almost no real stories.


4. Awkward Transitions


Sometimes, AI stitches paragraphs together mechanically, making transitions feel clunky or overly formal.


5. Long, “Safe” Sentences


To avoid mistakes, AI often writes overlong sentences that hedge meaning—playing it safe rather than being direct.


How AI Content Detection Is Changing


The upshot: AI text detection is getting more sophisticated.


  • Old tools hunted for obvious patterns (like repeated phrases or em dashes).
  • New methods analyze structure, vocabulary variety, and syntax—trying to spot “smoothness” and predictability.

For creators, this means old tricks to mask AI content won’t cut it. You need more complex tactics—including manual editing.


What This Means for AI Creators and Marketers


  • Raw AI output is risky. Just copying and pasting from ChatGPT isn’t safe anymore. You risk getting flagged and losing trust or visibility.
  • AI is a tool, not a replacement. Injecting personal insights, real cases, and unusual phrasing makes your work feel more human.
  • Editing for human touch is now crucial. The most successful AI creators blend automated drafts with hands-on refinement.

Practical Tips: How to Make AI Text Less Robotic


1. Add Personal Stories and Details


Drop in short case studies, anecdotes, or real-life comparisons—these are tough for AI to fake convincingly.


2. Use Unusual Expressions


Play with metaphors, humor, and offbeat comparisons. This breaks up the “smoothness” of AI text.


3. Check for “Evenness”


Mix up your sentence lengths and punctuation. Combine short, punchy lines with longer ones.


4. Leave Small “Imperfections”


A typo here, an unexpected word there, or an odd transition can add believability.


5. Use AI as a Drafting Tool


Let the AI produce the first draft. Then, revise it yourself—add emotion, examples, and direct questions for readers.


What to do right now


  • Review your recent AI-generated texts: Look for repetitive patterns and cut out overly “smooth” sections.
  • Add personal experiences and emotion to every piece, even if 90% of it is AI-generated.
  • Use AI for drafts only—never publish raw output without manual editing.
  • Vary sentence length and structure: Mix in short, abrupt sentences and surprising twists.
  • Test your content with AI detectors (like GPTZero, Copyleaks) to see how it holds up.

More hands-on AI tools and templates — in my courses.


Based on material from Hacker News