Undetectable AI detector

The detector is free and the pricing page lists "Unlimited AI Detecting" on every paid plan, so the plan table below prices the humanizer words that each plan sells, not detection (undetectable.ai/pricing, read 9 November 2026). The page's toggle defaults to Yearly with "50% off": each card shows the annual per-month price, the monthly price struck through, and the annual total. The footnote "Free Trial is limited to 250 words" applies to the humanizer. Figures are as displayed, read 9 November 2026.

The API meters detection against the same word balance as humanizing, at a tenth of the rate: "If you check 100 words using the Check for AI API, you will be deducted 10 credits from your word balance" (undetectable.ai/develop, read 9 November 2026).

Retention has two sources that say different amounts. The detector FAQ states: "Your text is never stored or used to train models" (undetectable.ai/ai-detector, read 9 November 2026), and the card's retention rank field takes that clause as a stated zero. The privacy policy, last updated 03/18/2023, does not mention submitted text and keeps personal information "for as long as it is necessary for the purposes set out in this privacy notice, unless a longer retention period is required or permitted by law" (privacy policy, read 9 November 2026). The terms add that the Services are hosted in the United States (terms of use, last updated 01/01/2025, read 9 November 2026).

The headline on the detector page reads: "The Most Accurate AI Checker: 99%+ Accuracy Proven By Independent Tests" (undetectable.ai/ai-detector, read 9 November 2026). The vendor states an accuracy rate of 99%+ and supports it with a comparison table of media reviews and one journal study. The table credits Undetectable AI with "100% detection / 0% false positives" from a study of ten free detectors published in the Indian Journal of Psychological Medicine, which the page summarises as detecting "100% of AI-generated text" and "0% false positives on human-written content" (same page).

That study is not yet a record in this index, so the 0% appears on the card as a vendor-stated figure with the source marked "vendor", the same handling the false positive ranking gives Pangram and Codequiry. The API documentation repeats the number in different words: the main result score is "the most accurate result, with 99%+ accuracy" (detector API documentation, read 9 November 2026). No sample size or test date accompanies either statement on the vendor's own pages.

The vendor's caveats are specific. At least 200 words are recommended for best accuracy, English offers the highest accuracy, and the per-tool scores are approximate (API documentation, read 9 November 2026). The pricing page describes the humanizer with a "money-back guarantee if output flagged is not human", and the terms refund credits only "if a submission provably flags as less than 75% human on a reputable AI detector" (terms of use, read 9 November 2026). A detector built by a vendor whose paid product is judged by whether it passes detectors is the caveat a reader should carry away.