How can I tell whether text was written by AI?

You cannot tell reliably, and no tool can prove it. OpenAI withdrew its own AI-text classifier in July 2023 for low accuracy: it caught 26% of AI-written text and wrongly flagged 9% of human writing. A Stanford study found detectors labelled 61% of essays by non-native English writers as AI-made, which matters for anyone writing in a second language, and a 2023 test of fourteen detectors found them neither accurate nor reliable, none reaching 80% accuracy. Turnitin puts its own false positive rate under 1% for documents with 20% or more AI writing, and around 4% for single sentences. Free checkers such as GPTZero or Quetext's detector (up to 1,000 words) can start a conversation, never settle one. Google's SynthID can confirm text came from the Gemini app, and only that. What works better: ask for drafts or version history, ask the writer to explain the piece, and check every fact and citation. Rule: treat a detector score as a question, not a verdict.