GPT Radar (Neuraltext)

One record in the studies index tested the tool. Walters 2023 ran GPT Radar between June 25 and July 12, 2023 on 42 ChatGPT-3.5 literature reviews, 42 ChatGPT-4 literature reviews and 42 human-written first-year composition papers. Table 4 prints 76% of the 126 documents correct, 24% incorrect and 0% uncertain; on the 84 AI papers 64% correct and 36% called human; on the GPT-3.5 set 98% correct and on the GPT-4 set 31% correct with 69% called human; and on the 42 human papers 100% correct with none called AI (DOI 10.1515/opis-2022-0158, Table 4, read 8 November 2026). The paper's text groups the result: "Crossplag, GPT Radar, and Writer each have false negative rates of 23-36%, while the other detectors have a maximum rate of 17% and a mean of 6.5%."

The fields follow the index rules. The false negative value is 36%, the printed share of AI papers called human, because the study record names GPT Radar in that range and defines the rate as the share of the 84 AI papers called human. The false positive value stays empty: Table 4 prints 0 of 42 human papers called AI (DOI 10.1515/opis-2022-0158, read 8 November 2026), but the study record prints a human-paper figure only for the detectors it names, and the rule for this wave takes a Walters false positive only where the record prints it. Both readings are written into the data file for the merge. GPT Radar was decisive where many tools hedged, with no uncertain verdict on any of the 126 documents, and that decisiveness cut both ways: every human paper passed and 69% of the GPT-4 papers passed too (Table 4).

GPT Radar ranks on one study record in independently tested detectors, at 36% in the false negative ranking, each read in the data file on 10 November 2026. The 400-word line is quoted in the minimum input ranking's second table as advice rather than ranked, because the page never enforced it (the reading the index gave GPTKit). It appears in the GPT-4 question at 31% of 42 GPT-4 papers and on which AI detectors were shut down, where its redirect is dated from the Wayback index. It does not appear in the free tier, input limit, language, retention or claim gap rankings, because the vendor published no quota, no maximum, no language, no period and no accuracy figure.

Every detector card on the register shows GPT Radar with the same fields as the live tools, so a reader who finds a 2023 GPT Radar verdict in an old editorial file can see what the one study measured and what the page promised.