An article discussing AI humanizer tools and the limits of detection software references ZeroGPT as a screening signal. This highlights the ongoing industry debate around the reliability and proper usage of AI detection software.
“Soul Food Project” podcast host Ian Yusuf found out about the YouTube channel “Potlikker” (formerly “The Hungry Historian”) from an unlikely source: an older relative. “I was out at dinner with some family friends. We were all celebrating,” Yusef tells TODAY.com. “I was talking to them about the podcast and everything, and an elder family friend texted me later and said, ‘Check out this soul food YouTube channel.’ It was The Hungry Historian.” Yusef says, when he got home, he visited the channel, which has since been renamed after a staple component of Black American cooking: Potlikker is the liquid left behind after cooking collard, mustard or turnip greens. Yusef, who is Black, looked through the page’s thumbnails and felt something off about the titles, like “Big Mama vs Betty Crocker: 25 Dishes No Cookbook Ever Printed,” as well as captions attempting African American Vernacular English (AAVE), like “preacher came we ate good” or “big mama tricks.” According to AI image and video detectors ZeroGPT and DeepAI, much of the content is AI-generated, including the thumbnails of elderly Black women, the “archival” clips in each video and even the main voiceovers. According to ElevenLabs’s AI Speech Classifier, voiceovers on Potlikker are using its Attank voice, which is described as a “middle-aged Male with African accent.” Everything on the channel aimed to highlight the culinary heritage of Black Americans — without using a single real person to do so. “The fact that it was about food bothered me, but really the fact that they’re preying on Black elders is really what tipped me over the edge,” he says. Using video-generation tools like Sora, image-generation tools like MidJourney and audio-generation tools like ElevenLabs, anyone with an internet connection can conjure up anything they like, which often leads to people pretending to be someone they aren’t. Yusef says that while he did not find out whether or not the page’s owners are Black, the only photo he found
An investigation revealed that a popular YouTube recipe channel relies entirely on AI-generated images, videos, and voiceovers. Tools like ZeroGPT helped identify the content as artificial.
Phrasly launched an AI humanizer model called Ultra that scores high on detection tools including Originality.ai. This highlights Originality.ai as a benchmark for AI text detection.
Commenters on an open-source article discuss whether an announcement was generated by AI using detection tools. This minor mention showcases the relevance of AI detection tools in verifying content origins.
This article discusses the ongoing debate regarding the reliability of AI detection tools in academic settings, highlighting instances where ZeroGPT and similar software produced false positives that led to unfair student investigations.
Headlines surfaced and scored by PortcoMonitor. Get the full, ranked feed + a daily briefing →