Signs of AI writing

Chatbots have habits, the way people have accents. Once you know them you see them everywhere: emails, reviews, LinkedIn, school essays.

Try the AI-o-meter

Just for fun: it only counts writing habits, nothing leaves your browser, and it is not an AI detector. Real detectors are unreliable too, so never accuse anyone on the strength of one.

Ten tells

  1. Em dashes — lots of them. Most people type a hyphen or a comma, or nothing.
  2. Favourite words: delve, tapestry, testament, realm, landscape, navigate, seamless, robust, foster, elevate, pivotal, vibrant.
  3. "It's not just X, it's Y." The chatbot's favourite way to sound profound.
  4. Things in threes: "clear, concise, and compelling".
  5. "It's important to note…" and other throat-clearing.
  6. A neat wrap-up: "In conclusion", "Overall", "Ultimately", even in a short message.
  7. Bullet points with bold lead-ins where a person would just write a sentence.
  8. Eager-assistant lines: "Great question!", "I hope this helps!", "Let me know if you'd like more".
  9. Even, polished paragraphs with no typos, no slang and no strong opinions.
  10. Specific-sounding but empty: lots of words, no detail you could check.

The same message, two ways

We wrote both versions of each to show the difference.

Asking for more time at work

Chatbot-style: "I hope this message finds you well. I wanted to reach out regarding the upcoming deadline. Due to unforeseen circumstances, I would greatly appreciate a brief extension to ensure the deliverable meets our high standards."

Person-style: "hey, the report's going to be late, sorry. the data from finance only came in this morning. friday ok?"

The person gives a reason you could check (finance, this morning) and a date. The chatbot gives "unforeseen circumstances" and "high standards".

A product review

Chatbot-style: "This kettle is a game-changer! Sleek, efficient, and easy to use, it seamlessly elevates my morning routine. Highly recommend!"

Person-style: "Boils fast. Lid is stiff, I need two hands. Lime scale after 3 weeks (hard water here). 4 stars."

A LinkedIn post

Chatbot-style: "Excited to share that I've embarked on a new chapter! This journey has been a testament to resilience, growth, and the power of community. It's not just a new role, it's a new mindset. 🚀"

Person-style: "New job! I'm joining the data team at a logistics company in Leeds. Mostly route planning. Nervous, mainly about the 7am stand-ups."

Where the tells show up

The tells are moving

Chatbots can be told to write casually, and people are learning to strip the obvious habits. As the famous tells fade, the lasting one is the same as in the hard round of the quiz: generated text reaches for the most likely words and the most common example. Real writing has the odd, checkable detail.

Before you accuse anyone

Every one of these is also something people do. Careful writers use em dashes; non-native speakers often write formally; some people love a bullet point. And AI detectors make mistakes too, including flagging real students' work. Use the tells as a reason to look closer, never as proof.

Spotting it in conversation is a different skill

In a chat, the giveaway is less the words and more the texture: no "now", no laziness, no real memories. That's what the daily game trains: three suspects, two humans, one AI, four questions.

Play today's round →

Questions

Is the AI-o-meter an AI detector?

No. It counts a handful of writing habits for fun. It runs in your browser and sends your text nowhere.

Why do chatbots use so many em dashes?

They learned from a lot of edited, published writing, where em dashes are common, and tend to overuse the patterns they learned.

Can AI writing be made undetectable?

Often, by asking the chatbot to write casually or by editing it. That is why habits are a hint, not proof.

More guides

All guides

Two of them are people. One is an AI.

Today's round is ready. Ask up to four questions, then point at the bot. Takes a minute, free, no account.

Play today's round →