Now Reading: AI Text Detectors Produce Unreliable Results, The Hindu Finds
-
01
AI Text Detectors Produce Unreliable Results, The Hindu Finds
AI text detectors can wrongly label human writing as machine-generated, while missing AI-produced work, The Hindu reports. In its test of 186 words from a 2010 Pulitzer Prize-winning foreign journalism feature, Pangram…
The Story in Brief
AI text detectors can wrongly label human writing as machine-generated, while missing AI-produced work, The Hindu reports. In its test of 186 words from a 2010 Pulitzer Prize-winning foreign journalism feature, Pangram and Quillbot classified the passage as entirely human-written, but ZeroGPT gave it a 49.9% AI score. Turnitin has acknowledged that its tool is not infallible, warning that unusual literary or scholarly styles can confuse it.
The tools are widely used by students, educators, businesses and content creators, with several offering paid rewriting services designed to make AI text appear more natural. Their reliability is under pressure as systems from OpenAI, Google and Anthropic produce more human-like writing. The Hindu also reports that AI accusations have affected authors, including American novelist Mia Ballard and Commonwealth Short Story Prize entrant Jamir Nazir.
The Indian Opinion
The lazy claim that a detector score proves cheating is unsafe, just as the claim that every polished passage is human is fanciful. Commercial tools have incentives to promote confidence, yet their results can vary sharply on the same text. Schools, publishers and employers should treat scores as prompts for a fuller review, not verdicts. The practical test is simple: will a decision-maker examine drafts, sources and the writer’s process before imposing a penalty?
Source: thehindu.com
This story was synthesised by AI from the source linked above.
