HUMAN OR AI · TYPO TRAP
Do typos mean a text is human or AI?
No. A spelling mistake is a clue about the text, not proof of its author. Try choosing a better follow-up instead.
Same meaning. Different spelling.
Which next step gives you more useful context?
What the typo actually changes
Both versions make the same claim. Fixing “teh” does not add a source, a timestamp, or a reliable identity. A human can write carefully. An AI can be instructed to write casually. The practical move is to ask about the meaning, then check how the answer fits the earlier exchange.
How to play and read your score
A round draws six different scenarios from twelve original cases. Choose one action per scenario. A linked, context-building question earns one point; treating spelling as proof earns none. The choice locks after your answer.
0–2: In this exercise, you often chose a spelling shortcut. Try asking about a detail already in the message. 3–4: You mixed surface cues with context checks. 5–6: You consistently chose context-building follow-ups. This score describes your choices in this round, not your ability to detect AI.
Can AI make deliberate spelling mistakes?
Yes: a request to imitate informal writing can include misspellings. That possibility is why an error cannot authenticate a human. Conversely, a clean sentence does not prove machine authorship.
Should I deliberately add typos to look human?
Adding errors supplies no evidence about where a text came from. If authorship matters, keep drafts, sources and revision history rather than trying to engineer a particular surface style.
Question design and source notes
Designed October 14, 2026. These twelve paired examples and their feedback were written with AI for this exercise. They are fictional modern chat snippets, not excerpts attributed to real people. Each pair holds its meaning constant while changing spelling or register. There are no hidden “human” source labels to guess here: your task is to select the stronger next question.
The exercise is our editorial design, not a validated assessment. This page is not an AI detector. It does not classify submitted text or connect you to another player.
Evidence and further play
- Human or Not FAQ: typing habits and limited context can make people sound artificial. Reviewed October 14, 2026.
- Jannai et al., Human or Not? (2023): a conversational Turing-style game; not validation of this exercise.
- Liang et al. (2023): detector bias across writing groups, a reason to avoid treating surface style as identity proof.