DetectArena

DetectArena is a crowdsourced benchmark for AI text detectors. Users submit text, compare tools such as GPTZero, Originality.ai and Pangram through blind pairwise voting, and help build public Elo rankings.

DetectArena is a benchmarking platform for AI text detectors. Rather than being a detector itself, it routes a piece of text through several existing detection services and lets people judge which one handled it better. The detectors named on the homepage are GPTZero, Originality.ai, Winston AI, Pangram, Sapling and ZeroGPT, and the results feed a public Elo-style ranking that reflects how the tools perform in the eyes of real users. The site offers four ways to work. Battle sends the text to two randomly chosen detectors whose names stay hidden, and the user votes for the one that seemed more accurate. Side by Side lets the user pick two specific detectors and compare their output directly. Solo runs a single detector and shows a segment-level breakdown of the text. Battle Royale sends the text to every detector at once and displays the overall consensus. An optional selector lets submitters label the text as General, Academic, Creative and Social Media, Marketing or Technical, which adds context to the comparison. The audience is broad. Educators and editors who rely on a detector can see how their tool stacks up against competitors before trusting its verdicts. Writers worried about false flags can test their own drafts against many services in one place. Detector vendors and researchers can look at head-to-head results on a shared leaderboard. The homepage also streams recent battles, showing the matchup, a text excerpt and each detector's score, which gives a quick sense of how often tools disagree. DetectArena is labelled Season 0 and carries a version marker of v0, so it is an early product. The site itself states that rankings give comparative context rather than definitive verdicts, which is a fair framing for a system built on crowd votes. Compared with running each detector's own site separately, it saves effort and removes brand bias through blind voting. It does not replace the detectors, and a sibling project, Avatar Arena, extends the same blind-benchmark idea to AI video avatars.