Sealed on Base.

{"canonicalPath":"/v/ai-detectors-can-prove-a-19nds","receipt":{"challengePrompt":"Have a specific detector report, vendor validation study, or school policy? Send it for a counter-check.","claim":"AI detectors can prove a student used ChatGPT with 99% accuracy.","confidence":"High","createdAt":"2026-06-18T09:00:01.669Z","kicker":"No detector can prove ChatGPT use from text alone at 99% accuracy.","paidAt":"2026-06-18T09:00:01.668Z","reasons":[{"body":"AI-writing detectors estimate whether text resembles machine output. They do not identify who wrote it or prove that ChatGPT was used. Even strong vendor scores are not forensic proof.","stance":"weakens","title":"Detector scores are not proof"},{"body":"Public evaluations and vendor notices show false positives and false negatives. OpenAI withdrew its own classifier because its accuracy was too low. Research has also found higher false-positive risk for some non-native,","stance":"weakens","title":"Public record shows real error rates"},{"body":"A “99% accurate” claim can hide the base-rate problem. In a large class or school, even a 1% false-positive rate can accuse innocent students. Accuracy also varies by text length, subject, editing, and paraphrasing.","stance":"context","title":"The 99% claim needs context"}],"score":12,"shareLine":"FactPage check: FALSE — AI detectors cannot prove a student used ChatGPT with 99% accuracy.","slug":"ai-detectors-can-prove-a-19nds","sources":[{"publisher":"OpenAI","title":"AI classifier withdrawal","url":"https://openai.com/index/new-ai-classifier-for-indicating-ai-written-text/","usedFor":"OpenAI said its AI text classifier was no longer available because of its low rate of accuracy."},{"publisher":"Patterns / Cell Press","title":"Detector bias study","url":"https://www.cell.com/patterns/fulltext/S2666-3899(23)00130-7","usedFor":"Study finding AI-text detectors can falsely flag writing by non-native English writers."},{"publisher":"Turnitin","title":"AI writing detection guidance","url":"https://www.turnitin.com/products/features/ai-writing-detection","usedFor":"Vendor material to inspect for stated limits, false-positive cautions, and guidance on using detector results."}],"summary":"The claim overstates what AI detectors can do. They can provide a signal, but public evidence does not support using them as 99%-accurate proof that a student used ChatGPT.","vaultCategory":"Tech","verdict":"FALSE"},"type":"factpage.receipt","version":"fp-receipt-v1"}