This is a free AI text detector from Buildly LLC. It scans text for passages written by ChatGPT, Claude, Gemini and similar, returning sentence-level highlights and a calibrated confidence score rather than a single verdict, and it will flag partially AI-written passages inside otherwise human text. Up to 10,000 characters per scan with no account, a browser extension, and a free API at 1,000 requests a day. Text is stated not to be permanently stored, never shared and never used for training.
The engineering choices are the better ones available. Sentence-level highlighting with a confidence score is far more defensible than a document-level percentage, because it shows a reader where the judgement came from instead of asking them to trust a number. Being genuinely free, accountless and rate-limited generously is unusual, and the privacy statement is specific rather than reassuring.
None of which fixes the underlying problem, and it needs stating plainly: **AI text detection does not work reliably enough for the decisions people make with it.** OpenAI withdrew its own classifier for poor accuracy. Independent testing has repeatedly found high false-positive rates on writing by non-native English speakers, on formulaic technical prose, and on text run through a grammar checker. The advertised 99.2% accuracy is not a figure any detector has demonstrated on adversarial or real-world text. Used to satisfy curiosity this is free and interesting; used to accuse a student or reject a freelancer it will produce confident false positives, and the cost of those falls entirely on the accused.
