Overview
Real or Fake Text (hosted at roft.io) is not a productivity app or a detector you run on your own writing — it is an interactive game and research instrument designed to measure how good people are at telling human-written text from machine-generated text. Visitors pick a category, read a short passage, and guess whether a computer or a human authored it. The project is framed as a public experiment in AI literacy: it surfaces how easily fluent, confident-sounding prose can be synthetic. Rather than flagging documents for plagiarism or "AI-ness," it trains human judgment by repetition and feedback. In the broader AI-detection landscape it serves an educational role, helping teachers, students, and curious readers calibrate their own instincts about what machine text feels like.
Key Features
- Multiple text categories: Short Stories (2,059 available), New York Times articles (2,194), Recipes (2,199), and Presidential Speeches (297).
- Start-Game flow: jump straight into a randomized playlist or choose a specific category to challenge.
- Turing-test style interaction: read a passage, then decide human vs. computer.
- Large, varied corpus: thousands of labeled examples spanning fiction, news, instructions, and political speech.
- No signup friction: the experience loads and plays immediately in the browser.
- Awareness framing: explicitly asks "How good are you at knowing when text has been written by a computer?"
Pros
- Free and instantly accessible with zero onboarding.
- Entertaining way to build intuition about AI-generated writing.
- Broad, diverse datasets across genres make the challenge non-trivial.
- Useful classroom or workshop icebreaker for AI-literacy education.
- Low stakes: no scoring penalties, accounts, or privacy tradeoffs.
Cons
- Not a detection tool — it cannot analyze your own documents.
- Feedback on "correct" answers depends on the underlying labels, which users cannot verify.
- Limited value for professionals who need actual AI-content verification.
- Category balance and corpus freshness are unclear from the page alone.
- No progress tracking, leaderboards, or explanatory feedback on why a passage was AI or human.
Who It's For
This tool is best for educators, students, journalists, and generally curious readers who want to sharpen their sense of what machine text looks like. It works well as a teaching aid in media-literacy lessons or as a quick, thought-provoking exercise. It is not intended for researchers or businesses that need to audit content at scale.
Verdict
Real or Fake Text is a clever, lightweight public experiment rather than a working detector. As an educational and awareness-building game it succeeds: it is free, fun, and genuinely humbling once you realize how often synthetic text fools you. Just don't expect it to replace a real AI-detection utility when you need to verify actual documents.