The underlying phenomenon is real, repeatedly reproduced, and independently documented by multiple parties who ran the test themselves.

Decrypt ran the Declaration's preamble through four detectors in October 2024 and reported that ZeroGPT said the Declaration of Independence text was 97.93% AI-generated, while Quillbot identified the same text as "Human-written 100%" and GPTZero gave it an 89% probability of being written by humans . Decrypt described using the same excerpt that data scientist Christopher Penn had used, and separately testing an excerpt from E.M. Forster's 1909 story "The Machine Stops" rewritten by ChatGPT . In that second test, ZeroGPT identified the genuine human "The Machine Stops" text as likely human at 4.27%, but also mistakenly labeled the AI-generated version as human-written at 6.35% , meaning the tool failed in both directions in the same sitting.

Independent repeats produced similar numbers. The Sentinel reported that Zero GPT determined the document was 97.75% AI-generated, while Copy Leaks and Writers.com said it was human-written, and noted that Zero GPT is the first link that appears when searching for AI detectors on Google . The Dallas Express extended the test to a different founding document: ZeroGPT reported that the 1836 Texas Declaration of Independence was 86.54% AI GPT, despite the document being signed by 59 delegates at Washington-on-the-Brazos in March 1836, nearly 190 years before the rise of ChatGPT . That outlet also recorded the earlier result and the reaction it drew: Decrypt's October 2024 finding of 97.93% prompted Christopher Penn, chief data scientist at Trust Insights, to call AI detectors a "joke," warning they were "unsophisticated and harmful," and to say that "these tools are being used to do things like disqualify students, putting them on academic probation or suspension" .

The specific 99.99% figure in this post is newer and thinner. It traces to a screenshot circulated in late November 2025: a Reddit user uploaded the 1776 text into an AI-detection tool and the detector declared the document "99.99 per cent AI-generated," after which screenshots spread quickly across social media . A social post from the same week names the tool as ZeroGPT . I could not retrieve the original Reddit thread, and no outlet I found reproduced the run under stated conditions.

The mechanism the caption describes is broadly supported by the original reporting on the earlier Constitution version of this story. Ars Technica's July 2023 piece explained that if the language in a piece of text is not surprising based on the model's training, the perplexity will be low, so the detector is more likely to classify it as AI-generated, and the Constitution's language is so ingrained in these models that they classify it as AI-generated, creating a false positive . GPTZero's founder told that outlet that "The US Constitution is a text fed repeatedly into the training data of many large language models" .

The broader reliability critique is supported by refereed work, not just anecdotes. Liang et al. found that seven detectors misclassified over half of a set of TOEFL essays as "AI-generated," with an average false positive rate of 61.22%, with all seven unanimously identifying 18 of the 91 essays as AI-authored . The published version states that the authors "exposed an alarming bias in GPT detectors against non-native English speakers: over half of the non-native English writing samples were misclassified as AI generated, while the accuracy for native samples remained near perfect" . Separately, a 2023 study by Weber-Wulff et al. evaluated 14 detection tools , and its authors reported accuracy in many studies of "only around 50% or slightly above" .

Vendor conduct corroborates the limitation. OpenAI's own announcement page now carries the note that "As of July 20, 2023, the AI classifier is no longer available due to its low rate of accuracy" . Turnitin's chief product officer, after launching with a sub-1% claim, conceded a higher figure: the tool has a higher false positive rate than the company originally asserted, and the company has not disclosed the new document-level false positive rate .

Finally, the tool most often implicated markets itself confidently while disclaiming in the same breath. Its product page advertises 98.5% accuracy for sentence-level detection, while also stating that "No AI detector is 100% accurate — machine learning predictions are probabilistic by nature" .