How to spot AI writing (and why it's getting harder)
There's no single giveaway. AI text has habits, and so do people, and in short answers they overlap a lot. Here's what tends to separate them, with real examples from the game.
Signs a person wrote it
- Small mistakes. Typos (“Ofcourse”), missing words, odd spacing, run-on sentences. Today's AI rarely makes them unless asked to.
- Specific, lived-in details. A named team, a hometown, a child's age, a café on a particular road. AI tends to invent tidier, more generic details.
- Tangents and uneven rhythm. People ramble, repeat a word, or end abruptly. AI answers are usually evenly paced.
- Jokes that are a bit strange. “It dropped from the sky” is more human than a neatly balanced punchline.
- Being wrong with confidence. A half-remembered anecdote or a slightly wrong definition is very human.
Signs an AI wrote it
- Polish everywhere. Clean grammar, parallel lists, items of similar length.
- Balanced, hedged takes. “It depends… here are both sides… I lean toward…”.
- Quotable endings. A neat closing line that sounds like a caption.
- Warm but generic memories. A childhood piano teacher, a camping trip, a parent pointing at stars.
- Apologetic refusals, though people write those too: in one round the human starts with “Unfortunately, I can't answer that.”
Examples from the game
“What is the meaning of life?”
Some say it's love and family, others say it's 42. I say you should probably figure this one out for yourself.
Probably something like: find a few people you love, do work you're proud of, and laugh as often as you can. Nobody hands you a better answer than that.
The human cracks a Hitchhiker's Guide joke ('42') and shrugs the question back at you. The AI gives a sincere, quotable answer.
“Do you get scared?”
Ofcourse! Everyone feels scared of something or the other. It is only the intensity that makes it different. I mean, I may be scared of heights, but I can chose to not let it get to me even when I am at a height, by say singing a rhyme and distracting my mind. Fear is in your head. So be fearless.
Of course I do. Mostly little things, like walking into a dark house or waiting for test results. I've learned that being brave isn't not being scared, it's just doing the thing anyway.
The human writes 'Ofcourse' and 'chose' for 'choose', and wanders into advice; the AI is tidy and ends on a familiar motto.
“What is heavier, 1 centimetre or 5 miles?”
Unfortunately, I can't answer that question. A centimetre and a mile are both units of measurement and therefore don't have an associated weight.
Trick question! A centimetre and five miles are both lengths, not weights, so neither one is heavier.
The human apologizes ('Unfortunately, I can't answer that'), which sounds like a chatbot; the AI answer is breezier. A classic trap.
“In the english language what is alliteration?”
Alliteration as I learned it in school from my second grade teacher is the repetition of consonant vowels sounds at the beginning of words and the end of sentences. An example of alliteration is the following "people are so perplexing when they are pushy."
It's when a bunch of words close together start with the same sound, like 'Peter Piper picked a peck of pickled peppers.' Writers use it to make a line catchy.
The human mentions a second-grade teacher and gets the definition slightly wrong ('end of sentences'); the AI uses the most famous example.
Don't trust detectors either
Automated “AI detectors” make the same mistakes people do, and flag real writing, especially from non-native speakers. Treat every sign above as a hint, not proof. The useful habit is to check the claim, not guess the author.