By Our Correspondent
National News – Researchers and fact-checkers have raised alarm over dubious artificial intelligence detection tools that wrongly label human-written content as AI-generated before offering paid services to “humanise” the text.
Experts say the emerging practice is spreading online and may be used to scam writers, students and institutions while undermining trust in legitimate AI verification systems.
Investigations by international fact-checking teams found that several AI-detection platforms, including tools identified as JustDone AI, TextGuard and Refinely, frequently produced inaccurate results.
The tools analysed genuine human-written texts in multiple languages and repeatedly classified them as AI-generated content, sometimes claiming up to 88 per cent AI involvement.
The misleading results often appeared alongside prompts encouraging users to pay a fee to remove the alleged AI traces.
In one example, a report about tensions between the United States and Iran was analysed by a detector which incorrectly labelled most of the article as AI-written before advertising a paid option promising “100% unique text.”
Experts say the scheme works by exploiting widespread concern about AI-generated misinformation.
As artificial intelligence tools become more common in journalism, education and public communication, many individuals and organisations rely on detection systems to verify authenticity.
However, researchers warn that some of these detectors may not perform real analysis at all.
Tests revealed that certain tools continued to generate similar AI-detection results even without internet access, suggesting the responses could be scripted rather than based on advanced algorithms.
Debora Weber-Wulff, a German academic who studies AI detection systems, described the services as deceptive schemes designed to sell rewriting tools that often produce awkward or meaningless sentences instead of genuinely improving the text.
The issue also raises broader concerns about misinformation. Analysts say unreliable detectors can be used to falsely accuse individuals, journalists or political groups of using AI to fabricate documents or messages.
Such tactics contribute to what experts call the “liar’s dividend,” a strategy where authentic information is dismissed as artificial.
Universities and fact-checking organisations have also warned that AI detection technologies still face significant limitations and cannot guarantee accuracy.
As a result, investigators increasingly rely on additional evidence such as metadata, digital watermarks and open-source verification methods when analysing potentially AI-generated content.
Experts urge users to treat AI detection results cautiously and avoid paying for services that promise guaranteed humanisation of text
