Instead of detecting AI, record and replay writing as it happens
The website "Human Wrote This" proposes a system where writing is recorded and replayed as it happens, rather than relying on AI detection tools to determine if content was written by a human or machine.
Background
- Most AI-text detectors (GPTZero, Originality.ai) try to classify a finished piece of text as human- or machine-written, but they are unreliable, easy to fool, and often penalize non-native speakers.
- This site (HumanWroteThis) proposes a completely different approach: instead of scanning the final text, it records the **process** of writing — every keystroke, pause, deletion, and revision — so the audience can replay how the text was actually created.
- Human writing is non-linear: we backtrack, fix typos, restructure sentences. LLMs generate text in one forward pass. A recorded "edit history" is therefore much harder for an AI to fake convincingly.
- The project was created by a group of researchers and builders (including from Stanford and the NYU AI & Democracy initiative) who want to restore trust in online discourse without relying on black-box detectors.
- Why it matters: As AI-generated content floods the web, proving that something was written by a human (and showing the authentic process) could become more valuable than trying to catch fakes after the fact.